CN104463356A

CN104463356A - Photovoltaic power generation power prediction method based on multi-dimension information artificial neural network algorithm

Info

Publication number: CN104463356A
Application number: CN201410704747.5A
Authority: CN
Inventors: 王坚敏; 韩志军; 周卫国; 金山红; 陈国恩; 赵仰东; 王跃强; 张磊; 王业平; 任志翔; 罗里志; 董成明; 陈春
Original assignee: Jiaxing Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Current assignee: Jiaxing Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Priority date: 2014-11-27
Filing date: 2014-11-27
Publication date: 2015-03-25

Abstract

A method for forecasting photovoltaic power generation based on a multidimensional information artificial neural network algorithm, which includes the following steps: 1) calculation of similarity: using daily characteristic data provided by the Meteorological Bureau and day types that are more vaguely described as vectors for similarity calculation; 2) Selection of similar days: starting from the closest historical day, reversely calculate the similarity value F _N between the j-th day and the i-th day day by day; select m days with the highest similarity in the most recent period or days with a similarity greater than 0.80m as 3) The establishment of the forecast model and the training of the model: use a three-layer BP neural network to establish different training samples, and continuously inject new test data, and perform BP on historical days and similar photovoltaic power generation samples. Neural network training; 4) Output of forecast results: select m similar days of the forecast day, and input similar day samples into the trained BP neural network; 5) Forecast error analysis: use root mean square error as the model evaluation method.

Description

A Method for Predicting Photovoltaic Power Generation Based on Multidimensional Information Artificial Neural Network Algorithm

技术领域 technical field

本发明涉及的是一种基于多维信息人工神经网络算法的光伏发电功率预测方法，主要是用来提高电力系统光伏发电预测精度。 The invention relates to a method for predicting photovoltaic power generation based on a multidimensional information artificial neural network algorithm, which is mainly used to improve the prediction accuracy of photovoltaic power generation in a power system. the

背景技术 Background technique

光伏发电功率预测是能量管理中急需解决的问题。光伏发电与风力发电一样,均属于波动性和间歇性电源,而且,各用户或小区使用的光伏电池种类及其安装位置随机性也大，光伏发电系统受光照强度和环境、温度等气候因素的影响,输出功率的变化具有不确定性,输出功率的扰动将有可能影响电网的稳定,因此,需要加强光伏发电功率预测的研究,预先获得光伏发电系统的日发电量曲线,从而协调电力系统制定发电计划,减少光伏发电的随机化问题对电力系统的影响。使用蓄电池来稳定光伏发电功率输出是一种可行的方法,但需要追加成本,而且废旧蓄电池还会导致环境污染。因此,需要对光伏系统的发电功率进行准确预测,以便了解大规模的太阳能光伏并网系统的发电运行特性以及与电网调度、电力负荷等的配合问题,这样有助于整个电力系统的规划和运行,从而减少光伏发电随机性对电力系统的影响,提高系统的安全稳定性。 Photovoltaic power prediction is an urgent problem in energy management. Photovoltaic power generation, like wind power generation, is a fluctuating and intermittent power supply. Moreover, the types of photovoltaic cells used by each user or community and their installation locations are also random. The change of output power is uncertain, and the disturbance of output power may affect the stability of the power grid. Therefore, it is necessary to strengthen the research on photovoltaic power generation prediction, obtain the daily power generation curve of photovoltaic power generation system in advance, and coordinate the power system to formulate Power generation plan, reducing the impact of randomization of photovoltaic power generation on the power system. Using batteries to stabilize the power output of photovoltaic power generation is a feasible method, but it requires additional costs, and waste batteries will also cause environmental pollution. Therefore, it is necessary to accurately predict the power generation of the photovoltaic system in order to understand the operating characteristics of the large-scale solar photovoltaic grid-connected system and the coordination with the grid dispatching and power load, which will help the planning and operation of the entire power system. , so as to reduce the influence of randomness of photovoltaic power generation on the power system and improve the security and stability of the system. the

发明内容 Contents of the invention

本发明的目的在于克服现有技术存在的不足，而提供一种通过基于多维信息人工神经网络算法来提高电力系统光伏发电在线实时预测的精度和准确性，充分考虑对光伏发电预测影响的多种因素，构造合适的神经网络模型，建立不同的训练样本，进行反复训练的基于多维信息人工神经网络算法的光伏发电功率预测方法。 The purpose of the present invention is to overcome the deficiencies in the prior art, and to provide a multi-dimensional information-based artificial neural network algorithm to improve the precision and accuracy of the online real-time prediction of photovoltaic power generation in power systems, and fully consider the various influences on photovoltaic power generation prediction. Factors, constructing a suitable neural network model, establishing different training samples, and performing repeated training is a photovoltaic power prediction method based on multi-dimensional information artificial neural network algorithm. the

本发明的目的是通过如下技术方案来完成的，一种基于多维信息人工神经网络算法的光伏发电功率预测方法，该光伏发电功率预测方法包括如下步骤： The object of the present invention is accomplished through the following technical solutions, a method for predicting photovoltaic power generation based on a multidimensional information artificial neural network algorithm, the method for predicting photovoltaic power generation includes the following steps:

1)相似度的计算：利用气象局提供的日特征数据，包括平均温度、最高温度、最低温度等，以及气象局提供的比较模糊描述日类型，包括晴天、多云、阴天、雨天等作为相似度计算的向量；所述日类型信息主要是根据历史数据的统计分析及经验将晴天、多云、阴天、雨天等日类型信息映射为0-1之间的一个日类型指数并作为预测模型的输入变量；对于模型中用到的原始数据可能带来的不利，对数据进行归一化处理； 1) Calculation of similarity: use the daily characteristic data provided by the Meteorological Bureau, including average temperature, maximum temperature, minimum temperature, etc., and the relatively vague description day types provided by the Meteorological Bureau, including sunny days, cloudy days, cloudy days, rainy days, etc. The vector of degree calculation; the day type information is mainly based on the statistical analysis of historical data and experience to map the day type information such as sunny, cloudy, cloudy, rainy days into a day type index between 0-1 and used as the prediction model Input variables; normalize the data for the possible disadvantages of the original data used in the model;

2)相似日的选择：从最临近历史日开始，逆向逐日计算第j日与第i日的相似度值F_N；选取最近一段时间中相似度最高的m日或者相似度大于0.80m日作为预测日的相似日，m的取值大于3，以满足训练网络有较强的外推性； 2) Selection of similar days: starting from the closest historical day, reversely calculate the similarity value F _N between the j-th day and the i-th day day by day; select m days with the highest similarity in the most recent period or days with a similarity greater than 0.80m as The value of m is greater than 3 for the similar day of the forecast day, so as to satisfy the strong extrapolation of the training network;

3)预测模型的确立和模型的训练：采用三层BP神经网络，包含输入层、输出层和隐含层，建立不同的训练样本,并不断注入新的测试数据,对历史日以及其相似光伏发电功率样本进行BP神经网络训练； 3) The establishment of the prediction model and the training of the model: a three-layer BP neural network is used, including the input layer, output layer and hidden layer, different training samples are established, and new test data are continuously injected. Generating power samples for BP neural network training;

4)预测结果的输出：选择预测日的m个相似日，将相似日样本输入训练好的BP神经网络，得出预测日的光伏电场发电功率； 4) Output of forecast results: select m similar days of the forecast day, input the samples of similar days into the trained BP neural network, and obtain the photovoltaic electric field power generation on the forecast day;

5)预测误差分析：为保证模型预测误差的平稳性，采用均方根误差作为模型评估方法； 5) Prediction error analysis: In order to ensure the stability of model prediction error, the root mean square error is used as the model evaluation method;

$RMSE RMSE = = \sqrt{{Σ Σ}_{i i = = 00}^{N N} \frac{{(({P P}_{f f}^{i i} - - {P P}_{a a}^{i i}))}^{22}}{N N}}$

式中：N为数据的个数；为预测值；为实际值。 In the formula: N is the number of data; is the predicted value; for the actual value.

2、根据权利要求1所述的基于多维信息人工神经网络算法的光伏发电功率预测方法，其特征在于： 2. The photovoltaic power generation power prediction method based on multidimensional information artificial neural network algorithm according to claim 1, characterized in that:

所述的步骤1)中，主要是利用气象局提供的日特征气象信息来挑选相似日； In the described step 1), it is mainly to select similar days by utilizing the day characteristic meteorological information provided by the Meteorological Bureau;

光伏发电功率的影响因素构成如下向量： The influencing factors of photovoltaic power generation constitute the following vector:

$Y Y = = [[\overset{&OverBar; &OverBar;}{T T},, {T T}_{min min},, {T T}_{max max} X x]]$

式中：为平均温度，T_min、T_max分别为最低温度和最高温度，X为日类型； In the formula: is the average temperature, T _min and T _max are the minimum temperature and maximum temperature respectively, and X is the day type;

所述的归一化公式是： The normalization formula described is:

${P P}_{n no} = = \frac{{p p}_{n no} - - {p p}_{min min}}{{p p}_{max max} - - {p p}_{min min}}$

${N N}_{n no} = = \frac{{n no}_{n no} - - {n no}_{min min}}{{n no}_{max max} - - {n no}_{min min}}$

式中：p_n,n_n——原始输入、目标数据； In the formula: p _n , n _n —— original input, target data;

p_min，p_max，n_min,n_max——p和n中的最小值和最大值； p _min , p _max , n _min , n _max - the minimum and maximum values of p and n;

P_n，N_n——归一化后的输入、目标数据 P _n , N _n —— normalized input and target data

影响向量： Influence vector:

Y_p＝[Y_p(1),Y_p(2),Y_p(3),Y_p(4)]^T Y _p ＝[Y _p(1) ,Y _p(2) ,Y _p(3) ,Y _p(4) ] ^T

Y_N＝[Y_N(1),Y_N(2),Y_N(3),Y_N(4)]^T Y _N ＝[Y _N(1) ,Y _N(2) ,Y _N(3) ,Y _N(4) ] ^T

Y_P与Y_N在第k个因素的关联系数为 The correlation coefficient between Y _P and Y _N at the kth factor is

${ϵ ϵ}_{N N} ((k k)) = = \frac{\underset{N N}{min min} \underset{k k}{min min} | | {Y Y}_{P P} ((k k)) - - {Y Y}_{N N} ((k k)) | | + + ρ ρ \underset{N N}{max max} \underset{k k}{max max} | | {Y Y}_{P P} ((k k)) - - {Y Y}_{N N} ((k k)) | |}{| | {Y Y}_{P P} ((k k)) - - {Y Y}_{N N} ((k k)) | | + + ρ ρ \underset{N N}{max max} \underset{k k}{max max} | | {Y Y}_{P P} ((k k)) - - {Y Y}_{N N} ((k k)) | |}$

式中：ρ是分辨系数，其值一般取0.5；综合各点的关联系数，定义整个Y_P与Y_N的相似度为 In the formula: ρ is the resolution coefficient, and its value is generally taken as 0.5; the correlation coefficient of each point is integrated, and the similarity between the whole Y _P and Y _N is defined as

${F f}_{N N} = = {Π Π}_{k k = = 11}^{44} {ϵ ϵ}_{N N} ((k k))$

采用这种连乘方式定义相似度，可简单、自动地识别主导因素，并解决各因素权重设定问题。 Using this multiplication method to define similarity can easily and automatically identify dominant factors and solve the problem of setting the weight of each factor. the

步骤3)中：三层BP神经网络节点的确定有： In step 3): the determination of the three-layer BP neural network node has:

a、输入层节点的确定： a. Determination of input layer nodes:

输入层节点对应于模型的输入变量，模型采用10个输入变量，这10个输入变量为相似日的光伏发电数据； The input layer nodes correspond to the input variables of the model, and the model uses 10 input variables, which are the photovoltaic power generation data of similar days;

b、输出层节点的确定： b. Determination of output layer nodes:

这是由要预测的内容决定的。本文的输出向量(目标向量)为预测日的发电功率，因此输出层节点个数为1； This is determined by what is being predicted. The output vector (target vector) of this paper is the power generation power of the forecast day, so the number of nodes in the output layer is 1;

c、隐含层及隐节点数的确定： c. Determination of the number of hidden layers and hidden nodes:

隐含层节点常用的经验计算公式： Commonly used empirical calculation formulas for hidden layer nodes:

$m m = = \sqrt{n no + + l l} + + α α$

m＝log2ⁿ m=log2 ⁿ

$m m = = \sqrt{nl nl}$

以上各式中m为隐层节点数，n为输入层节点数，l为输出节点数，α为1-10之间的常数。由于输入层节点为10点，输出层节点为1点，综合考虑以上各式，终确定隐层节点数为5。 In the above formulas, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output nodes, and α is a constant between 1-10. Since there are 10 nodes in the input layer and 1 node in the output layer, considering the above formulas, the number of nodes in the hidden layer is finally determined to be 5. the

本发明的有益效果是： The beneficial effects of the present invention are:

通过多维信息人工神经网络算法，充分考虑对光伏发电预测影响的多种因素，使输入样本的准确度更高，预测模型的建立更适合，达到对光伏发电功率预测精度提高的效果。有助于电力系统的调度和运行。 Through the multi-dimensional information artificial neural network algorithm, fully consider various factors that affect the prediction of photovoltaic power generation, so that the accuracy of the input samples is higher, the establishment of the prediction model is more suitable, and the effect of improving the prediction accuracy of photovoltaic power generation is achieved. Contribute to the scheduling and operation of the power system. the

附图说明： Description of drawings:

图1为光伏发电预测的流程图。 Figure 1 is a flowchart of photovoltaic power generation forecasting. the

图2为某光伏电站的实际预测结果图。 Fig. 2 is the actual prediction result map of a photovoltaic power station. the

具体实施方式 Detailed ways

下面将结合附图及具体实施例对本发明作详细的介绍：图1所示，本发明所述的一种基于多维信息人工神经网络算法的光伏发电功率预测方法，该光伏发电功率预测方法包括如下步骤： The present invention will be described in detail below in conjunction with accompanying drawings and specific embodiments: As shown in Fig. 1, a kind of photovoltaic power generation power forecasting method based on multidimensional information artificial neural network algorithm described in the present invention, this photovoltaic power generation power forecasting method comprises the following steps:

1)相似度的计算，利用气象局提供的日特征气象信息来挑选相似日。 1) Calculation of similarity, using the daily characteristic meteorological information provided by the Meteorological Bureau to select similar days. the

式中：为平均温度，T_min、T_max分别为最低温度和最高温度，X为日类型。 In the formula: is the average temperature, T _min and T _max are the minimum temperature and maximum temperature respectively, and X is the type of day.

气象局提供的日类型是比较模糊的描述:如晴天、晴天到多云、阴天、阴天有小雨、小雨转大雨等，首先根据历史发电量的统计分析及经验将晴天、多云、阴天、雨天等日类型信息映射为0-1之间的一个日类型指数作为预测模型的输入变量。为消除原始数据不同的单位变量、不同数量级的差异，以消除原始数据形式不同所带来的不利，对数据进行归一化处理。 The day type provided by the Meteorological Bureau is a relatively vague description: such as sunny, sunny to cloudy, cloudy, cloudy with light rain, light rain to heavy rain, etc. First, according to the statistical analysis and experience of historical power generation, the sunny, cloudy, cloudy, cloudy, The day type information such as rainy days is mapped to a day type index between 0 and 1 as the input variable of the prediction model. In order to eliminate the differences of different unit variables and different orders of magnitude of the original data, and to eliminate the disadvantages caused by the different forms of the original data, the data were normalized. the

归一化公式： Normalization formula:

影响向量： Influence vector:

式中：ρ是分辨系数，其值一般取0.5。综合各点的关联系数，定义整个Y_P与Y_N的相似度为 In the formula: ρ is the resolution coefficient, and its value is generally taken as 0.5. Based on the correlation coefficient of each point, the similarity between Y _P and Y _N is defined as

${F f}_{N N} = = {Π Π}_{k k = = 11}^{44} {ϵ ϵ}_{N N} ((k k))$

2)相似日的选择，从最临近历史日开始，逆向逐日计算第j日与第i日的相似度值F_N。选取最近一段时间中相似度最高的m日或者相似度F_N≥r(r为一定的数值，本文中取r＝0.80)的m日作为第N日的相似日。一般地，相似日个数m要大于3，以满足训练网络有较强的外推性。 2) The selection of similar days, starting from the closest historical day, reversely calculates the similarity value F _N between the jth day and the ith day day by day. Select the m day with the highest similarity in the recent period or the m day with the similarity F _N ≥ r (r is a certain value, r=0.80 in this paper) as the similar day of the Nth day. Generally, the number m of similar days should be greater than 3 to satisfy the strong extrapolation of the training network.

3)预测模型的确定和模型的训练；采用三层BP神经网络，包含输入层、输出层和隐含层。 3) The determination of the prediction model and the training of the model; a three-layer BP neural network is used, including an input layer, an output layer and a hidden layer. the

a、输入层节点的确定 a. Determination of input layer nodes

b、输出层节点的确定： b. Determination of output layer nodes:

这是由要预测的内容决定的；本发明所述的输出向量(目标向量)为预测日的发电功率，因此输出层节点个数为1； This is determined by the content to be predicted; the output vector (target vector) of the present invention is the power generation of the forecast day, so the output layer node number is 1;

$m m = = \sqrt{n no + + l l} + + α α$

m＝log2ⁿ m=log2 ⁿ

$m m = = \sqrt{nl nl}$

以上各式中m为隐层节点数，n为输入层节点数，l为输出节点数，α为1-10之间的常数。由于输入层节点为10点，输出层节点为1点，综合考虑以上各式，终确定隐层节点数为5； In the above formulas, m is the number of hidden layer nodes, n is the number of input layer nodes, l is the number of output nodes, and α is a constant between 1-10. Since there are 10 nodes in the input layer and 1 node in the output layer, considering the above formulas, the number of nodes in the hidden layer is finally determined to be 5;

模型的训练：以历史日实际数据作为输出，将该历史日m个相似日相同时间的数据作为输入，对历史日以及其相似光伏发电功率样本进行BP神经网络训练。针对光伏发电系统未来一段时间发电量的不确定性,对不同的天气、季节下的样本数据进行划分,考虑光伏发电和气象部门天气预报的相关性,建立不同的训练样本,并不断注入新的测试数据,以提高精度； Model training: The actual data of the historical day is used as the output, the data of m similar days of the historical day at the same time is used as the input, and the BP neural network is trained on the historical day and its similar photovoltaic power generation samples. Aiming at the uncertainty of the photovoltaic power generation system's power generation in the future, divide the sample data under different weather and seasons, consider the correlation between photovoltaic power generation and the weather forecast of the meteorological department, establish different training samples, and continuously inject new ones. Test data to improve accuracy;

4)预测结果输出，选择预测日的m个相似日，将相似日样本输入训练好的BP神经网络，得出预测日的光伏电场发电功率，见图2所示。 4) Forecast result output, select m similar days of the forecast day, input the similar day samples into the trained BP neural network, and obtain the photovoltaic electric field power generation power of the forecast day, as shown in Figure 2. the

5)预测误差分析，为保证模型预测误差的平稳性，采用均方根误差作为模型评估方法； 5) Prediction error analysis, in order to ensure the stability of the model prediction error, the root mean square error is used as the model evaluation method;

实施例： Example:

本发明从历史日中选取最合适的相似日，使输入样本坏数据率更低，有利于预测收敛值更接近实际值。 The invention selects the most suitable similar day from the historical days, so that the bad data rate of the input sample is lower, and it is beneficial for the prediction convergence value to be closer to the actual value. the

本发明所述光伏发电功率的影响因素构成如下向量： The influencing factors of photovoltaic power generation power described in the present invention form following vector:

将神经网络应用于系统发电功率的预测问题时，训练网络的原始数据中，不同的变量通常以不同的单位变量，数量级的差异也比较大，如发电量的变化范围在0到100之间，然而气温通常的变化范围则在-10到40之间。由神经元激活函数的特性可以知道，神经元的输出通常被限制在一定的范围内，大多数人工神经网络的应用中使用的非线性激活函数为S函数，其输出被限定在(0，1)或(-1，1)之间，直接以原始数据对网络进行训练会引起神经元饱和，因此在对网络进行训练之前必须对数据进行预处理，以消除原始数据形式不同所带来的不利，通常的做法是归一化处理。研究表明：以恰当的方式对数据进行归一化处理可以加速神经网络的收敛。 When the neural network is applied to the prediction problem of system power generation, in the original data of the training network, different variables are usually variable in different units, and the magnitude of the difference is relatively large. For example, the variation range of power generation is between 0 and 100. However, the temperature usually varies from -10 to 40 degrees. It can be known from the characteristics of the neuron activation function that the output of the neuron is usually limited within a certain range. The non-linear activation function used in the application of most artificial neural networks is the S function, and its output is limited to (0, 1 ) or (-1, 1), training the network directly with the original data will cause neuron saturation, so the data must be preprocessed before the network is trained to eliminate the disadvantages caused by the different forms of the original data , the usual practice is to normalize. Studies have shown that normalizing data in an appropriate way can speed up the convergence of neural networks. the

归一化可以在模型的单个输入变量通道上独立进行，也可以对所有的输入通道一起进行。输入变量常用的归一化方法有以下几种：(1)简单归一化；(2)线性变换到[0，1]区间；(3)线性变换到区间[a，b]上。当需要输入和目标数据落入[0，1]区间时，归一化公式为 Normalization can be performed independently on a single input variable channel of the model, or it can be performed on all input channels together. Commonly used normalization methods for input variables are as follows: (1) simple normalization; (2) linear transformation to [0, 1] interval; (3) linear transformation to the interval [a, b]. When the input and target data need to fall into the [0, 1] interval, the normalization formula is

影响向量： Influence vector:

${F f}_{N N} = = {Π Π}_{k k = = 11}^{44} {ϵ ϵ}_{N N} ((k k))$

选择第N日相似日的具体步骤为： The specific steps for selecting a similar day on the Nth day are:

1)从最临近历史日开始，逆向逐日计算第j日与第i日的相似度值F_N。 1) Starting from the nearest historical day, reversely calculate the similarity value F _N between the jth day and the ith day day by day.

2)选取最近一段时间中相似度最高的m日或者相似度F_N≥r(r为一定的数值，本文中取r＝0.80)的m日作为第N日的相似日。 2) Select the m day with the highest similarity in the recent period or the m day with the similarity F _N ≥ r (r is a certain value, r=0.80 in this paper) as the similar day of the Nth day.

采用三层BP神经网络，考虑光伏发电和气象部门天气预报的相关性,建立不同的训练样本,并不断注入新的测试数据,对预测模型进行反复训练。 Using a three-layer BP neural network, considering the correlation between photovoltaic power generation and the weather forecast of the meteorological department, different training samples are established, and new test data are continuously injected to train the prediction model repeatedly. the

选择预测日的m个相似日，将相似日样本输入训练好的BP神经网络，得出预测日的光伏电场发电功率。 Select m similar days of the forecast day, input the samples of similar days into the trained BP neural network, and obtain the power generation power of the photovoltaic electric field on the forecast day. the

采用均方根误差作为模型评估方法。 The root mean square error was used as the model evaluation method. the

Claims

1., based on a photovoltaic power generation power prediction method for multidimensional information artificial neural network algorithm, it is characterized in that this photovoltaic power generation power prediction method comprises the steps:

1) calculating of similarity: the day characteristic utilizing weather bureau to provide, comprise medial temperature, maximum temperature, minimum temperature etc., and the comparison vague description day type that weather bureau provides, comprise the vector as Similarity Measure such as fine day, cloudy, cloudy, rainy day; Described day, the day such as fine day, cloudy, cloudy, rainy day type information to be mainly mapped as the day index of type as the input variable of forecast model between 0-1 according to the statistical study of historical data and experience by type information; For the raw data used in model may bring unfavorable, data are normalized;

2) selection of similar day: from closing on most history day, the reverse Similarity value F calculating jth day and i-th day day by day _n; Choose m day that in nearest a period of time, similarity is the highest or similarity is greater than the similar day of 0.80m day as prediction day, the value of m is greater than 3, has stronger extrapolation to meet training network;

3) establishment of forecast model and the training of model: adopt three layers of BP neural network, comprise input layer, output layer and hidden layer, set up different training samples, and constantly inject new test data, BP neural metwork training is carried out to history day and its similar photovoltaic generation power sample;

4) output predicted the outcome: m the similar day selecting prediction day, by the BP neural network that the input of similar day sample trains, draws the photovoltaic DC field generated output of prediction day;

5) predicated error analysis: for ensureing the stationarity of model predictive error, adopts root-mean-square error as model evaluation method;

RMSE = \sqrt{Σ_{i = 0}^{N} \frac{{(P_{f}^{i} - P_{a}^{i})}^{2}}{N}}

In formula: N is the number of data; for predicted value; for actual value.

2. the photovoltaic power generation power prediction method based on multidimensional information artificial neural network algorithm according to claim 1, is characterized in that:

Described step 1) in, mainly utilize weather bureau to provide day feature weather information to select similar day;

The influence factor of photovoltaic generation power is constructed as follows vector:

Y = [\overset{&OverBar;}{T}, T_{\min}, T_{\max}, X]

In formula: for medial temperature, T _min, T _maxbe respectively minimum temperature and maximum temperature, X is day type;

Described normalization formula is:

P_{n} = \frac{p_{n} - p_{\min}}{p_{\max} - p_{\min}}

N_{n} = \frac{n_{n} - n_{\min}}{n_{\max} - n_{\min}}

In formula: p _n, n _n---original input, target data;

P _min, p _max, n _min, n _max---the minimum value in p and n and maximal value;

P _n, N _n---the input after normalization, target data

Impact vector:

Y_{p} = {[Y_{p (1)}, Y_{p (2)}, Y_{p (3)}, Y_{p (4)}]}^{T}

Y_{N} = {[Y_{N (1)}, Y_{N (2)}, Y_{N (3)}, Y_{N (4)}]}^{T}

Y _pwith Y _nat the correlation coefficient of a kth factor be

ϵ_{N} (k) = \frac{\min_{N} \min_{k} | Y_{P} (k) - Y_{N} (k) | + ρ \max_{N} \max_{k} | Y_{P} (k) - Y_{N} (k) |}{| Y_{P} (k) - Y_{N} (k) | + ρ \max_{N} \max_{k} | Y_{P} (k) - Y_{N} (k) |}

In formula: ρ is resolution ratio, its value generally gets 0.5; The correlation coefficient of comprehensive each point, defines whole Y _pwith Y _nsimilarity be

F_{N} = Π_{k = 1}^{4} ϵ_{N} (k)

Adopt this company to take advantage of mode to define similarity, simply, automatically can identify leading factor, and solve each factor weight setting problem.

Step 3) in: defining of three layers of BP neural network node:

The determination of a, input layer:

Input layer corresponds to the input variable of model, and model adopts 10 input variables, and these 10 input variables are the photovoltaic generation data of similar day;

The determination of b, output layer node:

This is determined by the content that will predict.Output vector (object vector) is herein the generated output of prediction day, and therefore output layer node number is 1;

The determination of c, hidden layer and Hidden nodes:

The empirical formula that hidden layer node is conventional:

m = \sqrt{n + l} + α

m＝log2 ⁿ

m = \sqrt{nl}

Various middle m is the number of hidden nodes above, and n is input layer number, and l is output node number, and α is the constant between 1-10.Because input layer is 10 points, output layer node is 1 point, considers above various, determines that the number of hidden nodes is 5 eventually.