WO2022095265A1 - 一种基于改进平衡优化器算法的光伏电池参数识别方法 - Google Patents
一种基于改进平衡优化器算法的光伏电池参数识别方法 Download PDFInfo
- Publication number
- WO2022095265A1 WO2022095265A1 PCT/CN2020/141889 CN2020141889W WO2022095265A1 WO 2022095265 A1 WO2022095265 A1 WO 2022095265A1 CN 2020141889 W CN2020141889 W CN 2020141889W WO 2022095265 A1 WO2022095265 A1 WO 2022095265A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- cell
- ieo
- output
- layer
- parameters
- Prior art date
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 210000004027 cell Anatomy 0.000 claims description 66
- 230000006870 function Effects 0.000 claims description 43
- 239000002245 particle Substances 0.000 claims description 43
- 210000002569 neuron Anatomy 0.000 claims description 21
- 238000005457 optimization Methods 0.000 claims description 16
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000000694 effects Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 238000010248 power generation Methods 0.000 claims description 3
- 230000002028 premature Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 241000282461 Canis lupus Species 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000002803 fossil fuel Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
- H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/25—Design optimisation, verification or simulation using particle-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0499—Feedforward networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Definitions
- the invention belongs to the technical field of photovoltaic cells, and in particular relates to a photovoltaic cell parameter identification method based on an improved balance optimizer algorithm.
- the analytical method mainly relies on some key points on the I-V curve provided by the manufacturer, and based on this, the model is identified by a series of mathematical calculations. These methods are simple in structure, but the recognition accuracy is not high enough under the changing operating environment conditions. Deterministic methods mainly use the convexity of the P-V characteristic curve of the PV system for parameter identification, which can obtain relatively more accurate identification results, but these methods have extremely strict requirements on model characteristics and are highly sensitive to initial conditions and gradient information. This makes it easy for them to get stuck in local optima. Therefore, how to accurately and quickly identify the parameters of the PV cell model with limited measurement data has become the focus and difficulty of current research.
- the technical problem to be solved by the present invention is to provide a photovoltaic cell parameter identification method based on an improved balance optimizer algorithm, so as to solve the technical problems that the prior art cannot achieve the optimal parameter identification and easily fall into local optimum.
- a photovoltaic cell parameter identification method based on an improved balance optimizer algorithm comprising:
- Step 1 Establish PV cell model and fitness function
- Step 2 Based on the measured output I-V data, predict PV cells through BP neural network
- Step 3 Use the IEO algorithm to identify the parameters of the PV cell until the convergence condition of the IEO algorithm is reached, and finally output the identified optimal parameters.
- the method for establishing a PV cell model and a fitness function described in step 1 includes:
- IL and VL represent the output current and output voltage of the PV cell, respectively;
- I sh represents the current flowing through the parallel resistor R sh ;
- thermal voltage VT is defined as:
- T represents the battery temperature
- I ph is the photo-generated current
- I sd1 and I sd2 are the diode's Reverse saturation current
- R s is series resistance
- R sh is parallel resistance
- diode ideality factors a 1 , a 2 are parameters to be identified;
- the root mean square error RMSE is selected as the objective function.
- the objective function is required to be the smallest.
- the objective function formula is established as:
- RMSE(x) is the objective function
- RMSE represents the root mean square error
- N represents the number of experiments
- f( VL , IL ,x) is the error function expressed as:
- the fitness function is designed as an extension function of RMSE, and the expression is:
- N p represents the number of prediction data
- the method of predicting the output data of PV cells through the BP neural network based on the measured output I-V data described in step 2 is as follows: the main structure of the BP neural network is the same as that of the ordinary neural network. layer, hidden layer and output layer; the basic process includes two processes of forward propagation and back propagation; weights and biases are successfully updated based on the back propagation from the output layer to the input layer, and the above process is repeated until the final iteration is satisfied up to the standard.
- the method for predicting the output data of the PV battery through the BP neural network specifically includes:
- n represents the number of neurons in the (l-1) layer; represents the weight of the ith neuron in layer (l-1) pointing to the jth neuron in layer (l-1); represents the bias of the jth neuron in the lth layer; Representation is based on the Sigmoid function computational value, expressed as
- the error Lm for a single sample is defined as
- L is the loss function, is the output value of forward sample propagation;
- y m is the sample value;
- bias The partial derivative of can be calculated by:
- the weight matrix and bias matrix can be updated as follows
- ⁇ l and represents the weight matrix and its partial derivatives from the (l-1)th hidden layer to the lth hidden layer
- b l and represent the bias matrix and its partial derivative from the (l-1)th hidden layer to the lth hidden layer, respectively;
- Weights and biases can be successfully updated based on backpropagation from the output layer to the input layer, and the above process needs to be repeated until the final iteration criteria are met.
- step 3 using the IEO algorithm to identify the parameters of the PV cell until the convergence condition of the IEO algorithm is reached, and finally the method of outputting the identified optimal parameters includes:
- Step 3.1 the initialization phase, according to the following formula:
- C max and C min represent the minimum and maximum dimension values, respectively;
- ri represents a random vector between 0 and 1;
- Step 3.2 The solution vector of each PV cell is regarded as the optimization variable of the IEO; PV cell parameters are usually distributed within a certain range, so the optimization variable is limited within the upper and lower bounds, as follows:
- x h represents the h-th optimization variable (that is, the h-th PV cell parameter); Represents the upper and lower bounds of the hth variable; H represents the set of optimization variables;
- Step 3.3 calculate the fitness value of all particles according to the fitness function
- Step 3.4 build a balance pool according to the fitness values of all particles
- Step 3.5 calculate the selection probability of all candidate balanced particles
- Step 3.6 select a candidate balancing individual from the current balancing pool according to the selection probability
- Step 3.7 calculate the exponential term of the particle
- Step 3.8 calculate the particle generation rate
- Step 3.9 update the particle solution
- the particle's solution is updated by the following formula:
- Step 3.11 Determine whether the IEO stops iterating and converges.
- the method for constructing a balanced pool according to the fitness values of all particles described in step 3.4 is:
- the final convergent state is called the equilibrium state, containing the four best particles so far and another single particle, which are defined as equilibrium candidates, from which an equilibrium pool is generated:
- step 3.5 The method for calculating the selection probability of all candidate balanced particles described in step 3.5 is:
- IEO assigns different selection probabilities to all balancing candidates, and the fitness values of all balancing candidates are normalized to a range of 0 to 1, so the selection probability is given by The formula is determined:
- p min and p max represent the maximum and minimum selection probabilities
- p m represents the selection probability of the mth balance candidate
- E p represents the set of balance candidates.
- step 3.7 The method for calculating the exponential term of the particle described in step 3.7 is:
- time t represents an iterative function, expressed as follows:
- a 1 represents a constant value proportional to the global search capability
- step 3.8 The method for calculating the particle generation rate described in step 3.8 is:
- the generation rate G is as follows:
- r 1 and r 2 represent random values between 0 and 1, respectively;
- GCP is defined as the control parameter of the power generation rate;
- GP is the generation probability, which represents the proportion of particles that use generation to update the state.
- step 3.11 The method described in step 3.11 to determine whether the IEO stops iterating and converges is as follows:
- step 3.2 If k ⁇ k max , the IEO iteration ends, the algorithm converges, and the optimal solution is output, that is, the parameter identification result; otherwise, go back to step 3.2.
- the IEO algorithm adopted in the present invention predicts and expands the output I-V data samples through data prediction based on the BP neural network, thereby establishing a more reliable fitness function to effectively improve the solution quality.
- the present invention adopts the IEO algorithm to assign different selection probabilities to different balance candidates according to their fitness values, which can realize a more in-depth search, thereby improving the overall optimization efficiency.
- the actual performance of the IEO is comprehensively verified using a two-diode PV cell model.
- the case study shows that, compared with the grey wolf optimization (GWO) algorithm, the IEO algorithm in the present invention can effectively improve the optimization precision and efficiency.
- GWO grey wolf optimization
- Fig. 1 is a structure diagram of PV cell parameter identification based on IEO algorithm
- Figure 2 is a model diagram of a dual-diode PV cell
- Figure 3 is the structure diagram of BP neural network
- Figure 4a-d is a comparison of the PV cell output power-voltage (P-V) and I-V fitting characteristic curves identified by the IEO algorithm under 50% and 100% data;
- Figure 5a-b is a comparison chart of the convergence characteristics of IEO and GWO algorithm parameter identification under 50% and 100% data;
- Figure 6 is a comparison chart of the average error of parameter identification of IEO and GWO algorithms under different data sets.
- the present invention proposes a PV cell parameter identification method based on the IEO algorithm.
- the algorithm has fast convergence speed and high search accuracy, and can be used in PV cells. Efficient and accurate parameter identification in modeling simulation to achieve reliable PV cell modeling.
- the object of the present invention is achieved through the following technical solutions: first, establishing a PV cell model; secondly, using the BP neural network to predict the output data of the PV cell to establish a more reliable fitness function; then, using the IEO algorithm to identify the parameters of the PV cell , until the convergence condition of the IEO algorithm is reached, and finally the identified optimal parameters are output.
- the present invention adopts a common two-diode PV cell model, and the output current-voltage (I-V) characteristic equation is as follows:
- IL and VL represent the output current and output voltage of the PV cell, respectively;
- I sh represents the current flowing through the parallel resistor R sh ;
- thermal voltage VT is defined as:
- T represents the battery temperature
- the photo-generated current I ph , the reverse saturation currents I sd1 and I sd2 of the diode, the series resistance R s , the parallel resistance R sh and the diode ideality factors a 1 and a 2 are parameters that need to be identified.
- the main goal of parameter identification is to search for appropriate parameters to effectively reduce the errors between experimental and simulated data, which can be quantitatively evaluated by the objective function.
- the root mean square error (RMSE) is selected as the objective function.
- the objective function is required to be the smallest.
- the objective function formula is as follows:
- RMSE(x) is the objective function
- RMSE represents the root mean square error
- N represents the number of experiments
- f( VL , IL , x) is the error function, which can be expressed as:
- the objective function is directly selected as the fitness function in IEO.
- the fitness function can be designed as an extension function of RMSE, as follows:
- N p represents the number of prediction data
- the output data of the PV cell is predicted by the BP neural network, and the steps are as follows:
- the main structure of the BP neural network is the same as that of the ordinary neural network. It consists of three parts, namely the input layer, the hidden layer and the output layer.
- the structure diagram is shown in Figure 3.
- the basic process includes forward propagation and back propagation. two processes.
- n represents the number of neurons in layer (l-1); represents the weight of the ith neuron in layer (l-1) pointing to the jth neuron in layer (l-1); represents the bias of the jth neuron in the lth layer;
- Representation is based on the Sigmoid function computational value, which can be expressed as
- the error Lm for a single sample is defined as
- L is the loss function, is the output value of forward sample propagation; y m is the sample value.
- bias The partial derivative of can be calculated by:
- the weight matrix and bias matrix can be updated as follows
- ⁇ l and represents the weight matrix and its partial derivatives from the (l-1)th hidden layer to the lth hidden layer
- b l and denote the bias matrix and its partial derivative from the (l-1)th hidden layer to the lth hidden layer, respectively.
- the weights and biases can be successfully updated based on the back-propagation from the output layer to the input layer, and the above two processes need to be repeated until the final iteration criteria are met.
- the IEO algorithm is used to identify the parameters of the PV cell, and the specific steps are as follows.
- the initialization phase is carried out according to the following formula:
- C max and C min represent the minimum and maximum dimension values, respectively;
- ri represents a random vector between 0 and 1;
- PV cell parameters are usually distributed within a certain range, so the optimization variables are limited to their upper and lower bounds, as follows:
- x h represents the h-th optimization variable (that is, the h-th PV cell parameter); Represents the upper and lower bounds of the hth variable; H represents the set of optimization variables.
- the final convergent state is called the equilibrium state, which contains the four best particles so far and another single particle, which are defined as equilibrium candidates, from which an equilibrium pool can be generated:
- IEO assigns different selection probabilities to all balancing candidates. For example, a balanced candidate with a smaller fitness value has a higher probability of being selected. Note that the fitness values of all balanced candidates are normalized to a range of 0 to 1, so the selection probability can be determined by:
- p min and p max represent the maximum and minimum selection probabilities
- p m represents the selection probability of the mth balance candidate
- E p represents the set of balance candidates.
- the exponential term (F) is crucial to balance the global search and the local search, and is expressed as:
- time t represents an iterative function, expressed as follows:
- a 1 represents a constant value proportional to the global search capability
- the generation rate (G) is the key to providing accurate solutions by improving the local search stage as follows:
- r 1 and r 2 represent random values between 0 and 1, respectively;
- GCP is defined as the control parameter of the power generation rate;
- GP is the generation probability, which represents the proportion of particles that use generation to update the state.
- the particle's solution is updated by the following formula:
- step (2) If k ⁇ k max , the IEO iteration ends, the algorithm converges, and the optimal solution (ie, the parameter identification result) is output; otherwise, return to step (2).
- the IEO algorithm proposed by the present invention and the traditional GWO algorithm respectively perform parameter identification under the dual-diode PV cell model.
- 50%, 60%, 70%, 80%, 90% and 100% of the data acquisition gradients were randomly set from 26 pairs of measured data.
- only IEO uses forecast data, while other systems use raw measured data.
- the experimental results show that compared with the GWO algorithm, the IEO algorithm can use less measured data to identify the unknown parameters of the two-diode cell model of the PV cell faster and more accurately.
- Table 1 shows the unknown parameter values and errors of the PV cell two-diode model identified by the IEO algorithm. The results show that the convergence error of the IEO algorithm is smaller than that of the GWO algorithm, that is, the parameter identification results are more accurate.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Neurology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Photovoltaic Devices (AREA)
Abstract
本发明公开了一种基于改进平衡优化器算法的光伏电池参数识别方法,它包括:步骤1、建立PV电池模型和适应度函数;步骤2、基于测量的输出I-V数据,通过BP神经网络预测PV电池的输出数据;步骤3、利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数;解决了现有技术不能达到参数识别最优和容易陷入局部最优等技术问题。
Description
本发明属于光伏电池技术领域,尤其涉及一种基于改进平衡优化器算法的光伏电池参数识别方法。
如今,环境的严重恶化,如森林砍伐和空气污染,传统化石燃料等不可再生资源的迅速枯竭,正威胁着世界的可持续发展。为了应对日益严峻的环境问题,长期的能源革命和能源结构转型已成为必然。其中,太阳能被认为是最有前途和最有效的替代能源之一,由于其安装方便,储量丰富且无污染排放,已经获得了广泛的应用。准确识别PV电池模型相关的电气参数是PV***性能分析、故障诊断和最大功率点跟踪(maximum power point tracking,MPPT)中最重要和关键的一步。在现有技术中主要存在两种参数识别方法,即分析法和确定性方法。分析法主要依赖于制造商提供的I‐V曲线上的一些关键点,并据此通过一系列数学计算对模型进行参数识别。这些方法结构简单,但在变化的运行环境条件下识别精度不够高。确定性方法主要利用PV***的P‐V特性曲线的凸性进行参数识别,可以获得相对更准确的识别结果,但这类方法对模型特性要求极其严格,并且对初始条件以及梯度信息高度敏感,这使得它们很容易陷入局部最优。因此,如何在测量数据有限的情况下准确快速地识别PV电池模型的参数,成为了当下研究的重点和难点。
发明内容
本发明要解决的技术问题:提供一种基于改进平衡优化器算法的光伏电池参数识别方法,以解决现有技术不能达到参数识别最优和容易陷入局部最优等技术问题。
本发明技术方案:
一种基于改进平衡优化器算法的光伏电池参数识别方法,它包括:
步骤1、建立PV电池模型和适应度函数;
步骤2、基于测量的输出I‐V数据,通过BP神经网络预测PV电池
的输出数据;
步骤3、利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数。
步骤1所述建立PV电池模型和适应度函数的方法包括:
采用双二极管PV电池模型,输出电流‐电压current‐voltage,建立I‐V特性方程如下:
式中I
L和V
L分别表示PV电池的输出电流和输出电压;I
sh表示流过并联电阻R
sh的电流;热电压V
T定义为:
式中T代表电池温度;K=1.38×10
‐23J/K表示玻尔兹曼常数;q=1.6×10
‐19C表示电子电荷;I
ph为光生电流,I
sd1、I
sd2为二极管的反向饱和电流,R
s为串联电阻,R
sh为并联电阻,二极管理想因子a
1、a
2是需要识别的参数;
选择均方根误差RMSE作为目标函数,为实现最优的效果要求目标函数最小,建立目标函数公式为:
式中RMSE(x)为目标函数,x={I
ph,I
sd1,I
sd2,R
s,R
sh,a
1,a
2}表示需要识别未知参数的解向量;RMSE表示均方根误差;N表示实验次数;f(V
L,I
L,x)为误差函数表示为:
将适应度函数设计为RMSE的扩展函数,表达式为:
式中N
p表示预测数据的数量。
步骤2所述基于测量的输出I‐V数据,通过BP神经网络预测PV电池的输出数据的方法为:BP神经网络的主要结构与普通神经网络的主要结构相同,它包含三个部分,即输入层,隐含层和输出层;基本过程包括前向传播和反向传播两个过程;权重和偏置以基于从输出层到输入层的反向传播而成功更新,重复上述过程直到满足最终迭代标准为止。
所述通过BP神经网络预测PV电池的输出数据的方法具体包括:
前向传播
式中n代表第(l‐1)层中的神经元数;
表示第(l‐1)层中的第i个神经元指向第(l‐1)层中的第j个神经元的权重;
表示第l层中第j个神经元的偏置;
表示基于Sigmoid函数
计算的
的值,表示为
反向传播
将单个样本的误差L
m定义为
将学习率定义为μ,权重矩阵和偏置矩阵可以如下更新
权重和偏置可以基于从输出层到输入层的反向传播而成功更新,并且需要重复上述过程,直到满足最终迭代标准为止。
步骤3所述利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数的方法包括:
步骤3.1、初始化阶段,根据如下公式进行:
步骤3.2、每个PV电池的解向量被视为IEO的优化变量;PV电池参数通常分布在一定范围内,因此优化变量被限制在上下限内,如下所示:
步骤3.3、根据适应度函数计算所有粒子的适应度值;
步骤3.4、根据所有粒子的适应度值构建平衡池;
步骤3.5、计算所有候选平衡粒子的选择概率;
步骤3.6、根据选择概率从现在的平衡池中选择一个候选平衡个体;
步骤3.7、计算粒子的指数项;
步骤3.8、计算粒子的生成率;
步骤3.9、更新粒子的解;
粒子的解通过如下公式更新:
步骤3.10、更新迭代次数k=k+1;
步骤3.11、判断IEO是否停止迭代而收敛。
步骤3.4所述根据所有粒子的适应度值构建平衡池的方法为:
将最终收敛态称为平衡态,包含四个到目前为止最好的粒子和另一个单独粒子,它们被定义为平衡候选粒子,据此产生一个平衡池:
步骤3.5所述计算所有候选平衡粒子的选择概率的方法为:
与从平衡池中随机选择平衡候选对象不同,IEO对所有平衡候选对象分配有不同的选择概率,所有平衡候选对象的适应度值都被归一化为0到1的范围,因此选择概率由下式确定:
式中p
min、p
max表示最大、最小的选择概率;p
m表示第m个平衡候选的选择概率;E
p表示平衡候选的集合。
步骤3.7所述计算粒子的指数项的方法为:
在浓度更新过程中,指数项F表示为:
式中iter、iter
max分别表示目前和最大的迭代次数;a
2为一个针对局部搜索调整所采用的恒量;
为了有效地避免过早收敛,还增加:
步骤3.8所述的计算粒子的生成率的方法为:
生成率G具体如下:
步骤3.11所述判断IEO是否停止迭代而收敛的方法为:
如果k≥k
max,则IEO迭代结束,算法收敛,输出最优解,即参数识别结果;否则返回步骤步骤3.2。
本发明的有益效果:
1.本发明所采用的IEO算法通过基于BP神经网络的数据预测对输出I‐V数据样本进行预测与扩展,从而建立更可靠的适应度函数来有效提高求解质量。
2.本发明采用IEO算法根据不同的平衡候选对象的适应度值对其分配不同的选择概率,可以实现更深入的搜索,进而提高整体优化效率。
3.采用双二极管PV电池模型对IEO的实际性能进行了综合验证。实例研究表明,与灰狼优化(grey wolf optimization,GWO)算法相比,本发明采用IEO算法能有效提高优化精度和效率。
解决了现有技术不能达到参数识别最优和容易陷入局部最优等技术问题。
图1是基于IEO算法的PV电池参数识别结构图;
图2是在双二极管PV电池模型图;
图3是BP神经网络结构图
图4a‐d是50%和100%数据下IEO算法识别的PV电池输出功率‐电压(power–voltage,P‐V)和I‐V拟合特性曲线对比图;
图5a‐b是50%和100%数据下IEO和GWO算法参数识别的收敛特性对比图;
图6是不同数据集下IEO和GWO算法参数识别的平均误差对比图。
本发明为解决PV电池建模过程中其参数难以准确识别,影响精确建模的问题,提出一种基于IEO算法的PV电池参数识别方法,该算法 收敛速度快、搜索精度高,可在PV电池建模仿真中实现高效、精准的参数识别,从而实现可靠的PV电池建模。
本发明的目的通过以下技术方案实现:首先,建立PV电池模型;其次,利用BP神经网络预测PV电池的输出数据,建立更加可靠的适应度函数;然后,利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数。
具体步骤为:
建立PV电池模型(见附图2)和适应度函数,方法如下:
本发明采用常见的双二极管PV电池模型,输出电流‐电压(current‐voltage,I‐V)特性方程如下:
式中I
L和V
L分别表示PV电池的输出电流和输出电压;I
sh表示流过并联电阻R
sh的电流;而热电压V
T定义为:
其中T代表电池温度;K=1.38×10
‐23J/K表示玻尔兹曼常数;q=1.6×10
‐19C表示电子电荷。光生电流I
ph,二极管的反向饱和电流I
sd1、I
sd2,串联电阻R
s,并联电阻R
sh以及二极管理想因子a
1、a
2是需要识别的参数。
参数识别的主要目标是搜索适当的参数,以有效地减少实验数据和模拟数据之间的误差,这些误差可以通过目标函数进行定量评估。在此,选择均方根误差(RMSE)作为目标函数,为实现最优的效果,要求目标函数最小,其目标函数公式如下:
式中RMSE(x)为目标函数,x={I
ph,I
sd1,I
sd2,R
s,R
sh,a
1,a
2}表示需要识别未知参数的解向量;RMSE表示均方根误差;N表示实验次数;f(V
L,I
L,x)为误差函数,可表示为:
由于所有解都可以满足约束条件(3),因此在IEO中直接选择目标函数作为适应度函数。基于BP神经网络的预测输出数据,可以将适应度函数设计为RMSE的扩展函数,如下所示:
式中N
p表示预测数据的数量。
基于测量的输出I‐V数据,通过BP神经网络预测PV电池的输出数据,步骤如下:
BP神经网络的主要结构与普通神经网络的主要结构相同,它包含三个部分,即输入层,隐含层和输出层,结构图见附图3,其基本过程包括前向传播和反向传播两个过程。
(1)前向传播
其中n代表第(l‐1)层中的神经元数;
表示第(l‐1)层中的第i个神经元指向第(l‐1)层中的第j个神经元的权重;
表示第l层中第j个神经元的偏置;
表示基于Sigmoid函数
计算的
的值,可以表示为
(2)反向传播
将单个样本的误差L
m定义为
将学习率定义为μ,权重矩阵和偏置矩阵可以如下更新
因此,权重和偏置可以基于从输出层到输入层的反向传播而成功更新,并且需要重复上述两个过程,直到满足最终迭代标准为止。
所述利用IEO算法对PV电池的参数进行识别,具体步骤如下所述。
(1)初始化
初始化阶段,根据如下公式进行:
(2)优化变量
每个PV电池的解向量被视为IEO的优化变量。实际上,PV电池参数通常分布在一定范围内,因此优化变量被限制在其上下限内,如下所示:
(3)根据适应度函数(公式5)计算适应度值
(4)根据所有粒子的适应度值,构建平衡池
将最终收敛态称为平衡态,其中包含四个到目前为止最好的粒子和另一个单独粒子,它们被定义为平衡候选粒子,据此可以产生一个平衡池:
(5)计算所有候选平衡粒子的选择概率
与从平衡池中随机选择平衡候选对象不同,IEO对所有平衡候选对象分配有不同的选择概率。如,适应度值较小的平衡候选体有较高的被选择概率。注意,所有平衡候选对象的适应度值都被归一化为0到1的范围,因此选择概率可以由下式所确定:
式中p
min、p
max表示最大、最小的选择概率;p
m表示第m个平衡候选的选择概率;E
p表示平衡候选的集合。
(6)根据选择概率从现在的平衡池中选择一个候选平衡个体
(7)计算粒子的指数项
在浓度更新过程中,指数项(F)对于平衡全局搜索和局部搜索至关重要,表示为:
式中iter、iter
max分别表示目前和最大的迭代次数;a
2为一个针对局部搜索调整所采用的恒量。
为了有效地避免过早收敛,还考虑了:
(8)计算粒子的生成率
生成率(G)是通过改善局部搜索阶段提供准确解的关键,具体如下:
(9)更新粒子的解
粒子的解通过如下公式更新:
(10)更新迭代次数k=k+1
(11)判断IEO是否停止迭代而收敛
如果k≥k
max,则IEO迭代结束,算法收敛,输出最优解(即参数识别结果);否则,返回步骤(2)。
下面结合具体实例,对本发明作进一步说明。
本发明提出的IEO算法和传统GWO算法分别在双二极管PV电池模型下进行参数识别。在天气条件(光照强度G=1000W/m
2和温度T=33℃)下,从57mm的R.T.C.France太阳能电池中提取了用于模拟的总共26组测量I‐V数据集。为验证算法在实测数据不足情况下的优化性能,从26对实测数据中随机设置50%、60%、70%、80%、90%和100%的数据采集梯度。在这里,只有IEO使用预测数据,而其他***使用原始测量数据。
为了公平比较,IEO和GWO算法的最大迭代次数和种群规模是相同的,迭代均为100次,种群规模设计为50。每个算法在PV电池模型中独立运行100次。实验结果如下所示:
实施例:如图4的a‐d所示,无论是在50%数据集还是100%数据集下,利用IEO得到的模型曲线与实际数据具有高度的一致性,这也说明了它对于各种数据集下的PV电池参数识别具有突出的性能。由图5a和b所示,在50%和100%数据集下,IEO算法和GWO算法均能在20次以内收敛,但基于IEO算法的参数识别误差明显低于基于GWO算法的参数识别误差;且在两个数据集下IEO算法的收敛特性也展示出了出显著的稳定性。由图6可知,在测试的各个数据集下,与GWO算法相比,基于IEO算法的参数识别的平均误差更低。综上,实验结果表明:与GWO算法相比,IEO算法能够利用更少的实测数据更快、更准 地识别PV电池的双二极管电池模型的未知参数。表1展示了IEO算法所识别的PV电池双二极管模型的未知参数值及误差。结果表明,IEO算法相比于GWO算法收敛误差更小,也即参数识别结果更加精准。
表1 PV电池双二极管模型参数识别结果及误差
以上(结合附图)对本发明的具体实施方式作了详细说明,但是本发明并不限于上述实施方式,在本领域普通技术人员所具备的知识范围内,还可以在不脱离本发明宗旨的前提下做出各种变化。
Claims (10)
- 一种基于改进平衡优化器算法的光伏电池参数识别方法,它包括:步骤1、建立PV电池模型和适应度函数;步骤2、基于测量的输出I‐V数据,通过BP神经网络预测PV电池的输出数据;步骤3、利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数。
- 根据权利要求1所述的一种基于改进平衡优化器算法的光伏电池参数识别方法,其特征在于:步骤1所述建立PV电池模型和适应度函数的方法包括:采用双二极管PV电池模型,输出电流‐电压current‐voltage,建立I‐V特性方程如下:式中I L和V L分别表示PV电池的输出电流和输出电压;I sh表示流过并联电阻R sh的电流;热电压V T定义为:式中T代表电池温度;K=1.38×10 ‐23J/K表示玻尔兹曼常数;q=1.6×10 ‐19C表示电子电荷;I ph为光生电流,I sd1、I sd2为二极管的反向饱和电流,R s为串联电阻,R sh为并联电阻,二极管理想因子a 1、a 2是需要识别的参数;选择均方根误差RMSE作为目标函数,为实现最优的效果要求目标函数最小,建立目标函数公式为:式中RMSE(x)为目标函数,x={I ph,I sd1,I sd2,R s,R sh,a 1,a 2}表示需要识别未知参数的解向量;RMSE表示均方根误差;N表示实验次数;f(V L,I L,x)为误差函数表示为:将适应度函数设计为RMSE的扩展函数,表达式为:式中N p表示预测数据的数量。
- 根据权利要求1所述的一种基于改进平衡优化器算法的光伏电池参数识别方法,其特征在于:步骤2所述基于测量的输出I‐V数据,通过BP神经网络预测PV电池的输出数据的方法为:BP神经网络的主要结构与普通神经网络的主要结构相同,它包含三个部分,即输入层,隐含层和输出层;基本过程包括前向传播和反向传播两个过程;权重和偏置以基于从输出层到输入层的反向传播而成功更新,重复上述过程直到满足最终迭代标准为止。
- 根据权利要求3所述的一种基于改进平衡优化器算法的光伏电池参数识别方法,其特征在于:所述通过BP神经网络预测PV电池的输出数据的方法具体包括:前向传播式中n代表第(l‐1)层中的神经元数; 表示第(l‐1)层中的第i个神经元指向第(l‐1)层中的第j个神经元的权重; 表示第l层中第j个神经元的偏置; 表示基于Sigmoid函数 计算的 的值,表示为反向传播将单个样本的误差L m定义为将学习率定义为μ,权重矩阵和偏置矩阵可以如下更新权重和偏置可以基于从输出层到输入层的反向传播而成功更新,并且需要重复上述过程,直到满足最终迭代标准为止。
- 根据权利要求1所述的一种基于改进平衡优化器算法的光伏电池参数识别方法,其特征在于:步骤3所述利用IEO算法对PV电池的参数进行识别,直到达到IEO算法的收敛条件,最后输出识别的最优参数的方法包括:步骤3.1、初始化阶段,根据如下公式进行:步骤3.2、每个PV电池的解向量被视为IEO的优化变量;PV电池参数通常分布在一定范围内,因此优化变量被限制在上下限内,如下所示:步骤3.3、根据适应度函数计算所有粒子的适应度值;步骤3.4、根据所有粒子的适应度值构建平衡池;步骤3.5、计算所有候选平衡粒子的选择概率;步骤3.6、根据选择概率从现在的平衡池中选择一个候选平衡个体;步骤3.7、计算粒子的指数项;步骤3.8、计算粒子的生成率;步骤3.9、更新粒子的解;粒子的解通过如下公式更新:步骤3.10、更新迭代次数k=k+1;步骤3.11、判断IEO是否停止迭代而收敛。
- 根据权利要求5所述的一种基于改进平衡优化器算法的光伏电池参数识别方法,其特征在于:步骤3.11所述判断IEO是否停止迭代而收敛的方法为:如果k≥k max,则IEO迭代结束,算法收敛,输出最优解,即参数识别结果;否则返回步骤步骤3.2。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/786,556 US20230035108A1 (en) | 2020-11-09 | 2020-12-31 | Photovoltaic cell parameter identification method based on improved equilibrium optimizer algorithm |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011239017.4 | 2020-11-09 | ||
CN202011239017.4A CN112380765B (zh) | 2020-11-09 | 2020-11-09 | 一种基于改进平衡优化器算法的光伏电池参数识别方法 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022095265A1 true WO2022095265A1 (zh) | 2022-05-12 |
Family
ID=74578934
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/141889 WO2022095265A1 (zh) | 2020-11-09 | 2020-12-31 | 一种基于改进平衡优化器算法的光伏电池参数识别方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US20230035108A1 (zh) |
CN (1) | CN112380765B (zh) |
WO (1) | WO2022095265A1 (zh) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114970332A (zh) * | 2022-05-16 | 2022-08-30 | 西北工业大学 | 基于混沌量子麻雀搜索算法的锂电池模型参数辨识方法 |
CN115081483A (zh) * | 2022-07-01 | 2022-09-20 | 淮阴工学院 | 基于特征选择和gwo-bp的水轮发电机转子故障诊断方法 |
CN115221798A (zh) * | 2022-09-19 | 2022-10-21 | 中南大学 | 一种电池热过程时空建模预测方法、***、设备及介质 |
CN115986744A (zh) * | 2022-12-28 | 2023-04-18 | 国网安徽省电力有限公司芜湖供电公司 | 一种含共享储能的配电网潮流优化方法 |
CN116148680A (zh) * | 2023-04-20 | 2023-05-23 | 上海玫克生储能科技有限公司 | 电池参数的辨识方法、装置、计算机设备和存储介质 |
CN117494530A (zh) * | 2023-12-29 | 2024-02-02 | 湖北工业大学 | 一种基于改进灰狼算法的薄板钻孔夹具布局优化方法 |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113037213A (zh) * | 2021-03-04 | 2021-06-25 | 国网新疆电力有限公司信息通信公司 | 基于樽海鞘群算法的光伏电池模型参数辨识方法及装置 |
CN112711292B (zh) * | 2021-03-29 | 2021-07-09 | 深圳黑晶光电技术有限公司 | 一种光伏组件最大功率追踪方法、***及存储介质 |
CN116502989B (zh) * | 2023-06-27 | 2023-09-08 | 合肥城市云数据中心股份有限公司 | 一种基于混合平衡优化算法的冷链物流车辆路径优化方法 |
CN117114144B (zh) * | 2023-10-24 | 2024-01-26 | 青岛农业大学 | 基于人工智能的水稻耐盐碱性能预测方法及*** |
CN117451626B (zh) * | 2023-10-27 | 2024-05-28 | 清华大学 | 包含样品形状优化的叠层成像方法及装置 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120249047A1 (en) * | 2011-03-31 | 2012-10-04 | GM Global Technology Operations LLC | Photo-voltaic charging of high voltage traction batteries |
CN109992911A (zh) * | 2019-05-06 | 2019-07-09 | 福州大学 | 基于极限学习机和iv特性的光伏组件快速建模方法 |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5680324A (en) * | 1995-04-07 | 1997-10-21 | Schweitzer Engineering Laboratories, Inc. | Communications processor for electric power substations |
US10443139B2 (en) * | 2003-09-05 | 2019-10-15 | Brilliant Light Power, Inc. | Electrical power generation systems and methods regarding same |
CN103870649B (zh) * | 2014-03-19 | 2017-02-08 | 国家电网公司 | 一种基于分布式智能计算的主动配电网自治化仿真方法 |
CN109389236A (zh) * | 2017-08-07 | 2019-02-26 | 贵州电网有限责任公司电力科学研究院 | 电动汽车有序充电与无功优化协调控制策略 |
CN108054758B (zh) * | 2017-11-15 | 2020-06-19 | 清华大学 | 新能源电站电压平衡优化方法及存储介质 |
CN108562856A (zh) * | 2018-03-16 | 2018-09-21 | 江苏华西新能源工程技术有限公司 | 太阳能电池阵列并联内阻的在线检测装置及其检测方法 |
CN111191375B (zh) * | 2020-01-04 | 2023-06-02 | 温州大学 | 一种基于改进哈里斯鹰优化算法的光伏电池参数辨识方法 |
-
2020
- 2020-11-09 CN CN202011239017.4A patent/CN112380765B/zh active Active
- 2020-12-31 WO PCT/CN2020/141889 patent/WO2022095265A1/zh active Application Filing
- 2020-12-31 US US17/786,556 patent/US20230035108A1/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120249047A1 (en) * | 2011-03-31 | 2012-10-04 | GM Global Technology Operations LLC | Photo-voltaic charging of high voltage traction batteries |
CN109992911A (zh) * | 2019-05-06 | 2019-07-09 | 福州大学 | 基于极限学习机和iv特性的光伏组件快速建模方法 |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114970332A (zh) * | 2022-05-16 | 2022-08-30 | 西北工业大学 | 基于混沌量子麻雀搜索算法的锂电池模型参数辨识方法 |
CN114970332B (zh) * | 2022-05-16 | 2024-03-22 | 西北工业大学 | 基于混沌量子麻雀搜索算法的锂电池模型参数辨识方法 |
CN115081483A (zh) * | 2022-07-01 | 2022-09-20 | 淮阴工学院 | 基于特征选择和gwo-bp的水轮发电机转子故障诊断方法 |
CN115221798A (zh) * | 2022-09-19 | 2022-10-21 | 中南大学 | 一种电池热过程时空建模预测方法、***、设备及介质 |
CN115221798B (zh) * | 2022-09-19 | 2022-12-30 | 中南大学 | 一种电池热过程时空建模预测方法、***、设备及介质 |
CN115986744A (zh) * | 2022-12-28 | 2023-04-18 | 国网安徽省电力有限公司芜湖供电公司 | 一种含共享储能的配电网潮流优化方法 |
CN115986744B (zh) * | 2022-12-28 | 2024-02-06 | 国网安徽省电力有限公司芜湖供电公司 | 一种含共享储能的配电网潮流优化方法 |
CN116148680A (zh) * | 2023-04-20 | 2023-05-23 | 上海玫克生储能科技有限公司 | 电池参数的辨识方法、装置、计算机设备和存储介质 |
CN116148680B (zh) * | 2023-04-20 | 2023-06-23 | 上海玫克生储能科技有限公司 | 电池参数的辨识方法、装置、计算机设备和存储介质 |
CN117494530A (zh) * | 2023-12-29 | 2024-02-02 | 湖北工业大学 | 一种基于改进灰狼算法的薄板钻孔夹具布局优化方法 |
CN117494530B (zh) * | 2023-12-29 | 2024-03-26 | 湖北工业大学 | 一种基于改进灰狼算法的薄板钻孔夹具布局优化方法 |
Also Published As
Publication number | Publication date |
---|---|
CN112380765B (zh) | 2022-11-04 |
US20230035108A1 (en) | 2023-02-02 |
CN112380765A (zh) | 2021-02-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022095265A1 (zh) | 一种基于改进平衡优化器算法的光伏电池参数识别方法 | |
Wang et al. | Photovoltaic cell parameter estimation based on improved equilibrium optimizer algorithm | |
CN111860982A (zh) | 一种基于vmd-fcm-gru的风电场短期风电功率预测方法 | |
CN109978283B (zh) | 一种基于分支进化神经网络的光伏发电功率预测方法 | |
CN112116144B (zh) | 一种区域配电网短期负荷预测方法 | |
CN110942205B (zh) | 一种基于himvo-svm的短期光伏发电功率预测方法 | |
CN107562992B (zh) | 一种基于svm和粒子群算法的光伏列阵最大功率跟踪方法 | |
Bendali et al. | Deep learning using genetic algorithm optimization for short term solar irradiance forecasting | |
CN111814401A (zh) | 基于遗传算法的bp神经网络的led寿命预测方法 | |
CN114219181A (zh) | 基于迁移学习的风电功率概率预测方法 | |
Omer et al. | Comparison of decision tree based ensemble methods for prediction of photovoltaic maximum current | |
CN115995810A (zh) | 一种考虑气象波动自适应匹配的风电功率预测方法及*** | |
CN116345555A (zh) | 一种基于cnn-isca-lstm模型的短期光伏发电功率预测方法 | |
CN116632823A (zh) | 基于功率转换模型与多层感知机的短期光伏功率预测方法 | |
CN115759415A (zh) | 基于lstm-svr的用电需求预测方法 | |
CN115907122A (zh) | 区域电动汽车充电负荷预测方法 | |
CN116341717A (zh) | 一种基于误差补偿的风速预测方法 | |
CN114611757A (zh) | 基于遗传算法与改进深度残差网络的电力***短期负荷预测方法 | |
Jafarkazemi et al. | Performance prediction of flat-plate solar collectors using MLP and ANFIS | |
Smith et al. | Multi-objective evolutionary recurrent neural network ensemble for prediction of computational fluid dynamic simulations | |
CN115481788B (zh) | 相变储能***负荷预测方法及*** | |
CN114234392B (zh) | 一种基于改进pso-lstm的空调负荷精细预测方法 | |
Tan et al. | Long-term load forecasting based on feature fusion and lightgbm | |
Ban et al. | Electrical characteristics estimation of photovoltaic modules via cuckoo search—Relevant vector machine probabilistic model | |
Yalçın et al. | Prediction and Analysis of Weather Parameters with Global Horizontal Solar Irradiance Using LSTM-CNN Based Deep Learning Technique |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20960722 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20960722 Country of ref document: EP Kind code of ref document: A1 |