CN111367349A - Photovoltaic MPPT control method and system based on prediction model - Google Patents
Photovoltaic MPPT control method and system based on prediction model Download PDFInfo
- Publication number
- CN111367349A CN111367349A CN201811596118.XA CN201811596118A CN111367349A CN 111367349 A CN111367349 A CN 111367349A CN 201811596118 A CN201811596118 A CN 201811596118A CN 111367349 A CN111367349 A CN 111367349A
- Authority
- CN
- China
- Prior art keywords
- model
- prediction
- photovoltaic
- power point
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05F—SYSTEMS FOR REGULATING ELECTRIC OR MAGNETIC VARIABLES
- G05F1/00—Automatic systems in which deviations of an electric quantity from one or more predetermined values are detected at the output of the system and fed back to a device within the system to restore the detected quantity to its predetermined value or values, i.e. retroactive systems
- G05F1/66—Regulating electric power
- G05F1/67—Regulating electric power to the maximum power available from a generator, e.g. from solar cell
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/50—Photovoltaic [PV] energy
- Y02E10/56—Power conversion systems, e.g. maximum power point trackers
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Power Engineering (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Automation & Control Theory (AREA)
- Control Of Electrical Variables (AREA)
Abstract
The invention discloses a photovoltaic MPPT control method based on a prediction model, which comprises the following steps: acquiring input data of a prediction model in real time, wherein the input data of the prediction model comprises illumination intensity data and environment temperature data; determining a maximum power point voltage parameter predicted by each submodel in a preset combined prediction model according to input data of the prediction model, and further performing combined calculation on a prediction result of each submodel and a corresponding distribution weight according to the distribution weight corresponding to each submodel to obtain a maximum power point combined prediction voltage parameter; and generating a corresponding pulse control signal according to the current combined predicted voltage parameter, inputting the pulse control signal into a boosting driving circuit, and further controlling a grid-connected inverter to complete inversion processing so that the photovoltaic system operates at the maximum power point. The invention enables the photovoltaic array to stably operate at the maximum power point, shortens the time for tracking the maximum power point and has stronger adaptability in a complex environment.
Description
Technical Field
The invention relates to the technical field of photovoltaic power generation, in particular to a photovoltaic MPPT control method and system based on a prediction model.
Background
Among clean energy, solar energy is more and more concerned by people due to the characteristics of wide distribution, rich resources and the like, and photovoltaic power generation has huge potential and is bound to become a source of main energy in the world energy and electric power field in the future. The most critical technology for converting solar energy into electric energy is Maximum Power Point Tracking (MPPT), and the electric energy converted by the photovoltaic cell is output at the maximum power as much as possible by using a specific control method. However, the photovoltaic cell is affected by factors such as illumination radiation and temperature, so that the P-U characteristic of the photovoltaic cell fluctuates greatly, the maximum power of the system is tracked too frequently, and the system cannot stably operate at the maximum power point, thereby reducing the photovoltaic power generation efficiency.
At present, MPPT methods for photovoltaic power generation systems can be divided into two categories: traditional method, predictive method. The traditional methods mainly comprise a constant voltage tracking method, a short-circuit current proportionality coefficient method, a disturbance observation method, a conductance incremental method and the like. The constant voltage tracking method and the short-circuit current proportionality coefficient method are used for calculating the maximum power point voltage based on basic approximate rules of the output characteristics of the photovoltaic cell (for example, an approximate linear relation exists between the maximum power point voltage and the open-circuit voltage) so as to achieve the purpose of control. Although the method is simple and easy to implement, the method has large dependence on output characteristics and low efficiency. A disturbance observation method and a conductance incremental method are used for comparing power obtained by calculation according to detected values of output current and voltage of a photovoltaic cell and then carrying out maximum power tracking through voltage loop closed-loop control, and the method also belongs to a self-optimization algorithm. In addition, the prediction method is used for modeling a large amount of historical data by applying the mathematical statistics idea, comprises the steps of predicting by methods such as classical mathematics and artificial intelligence, obtaining a voltage value corresponding to the maximum power point through prediction, calculating to obtain a PWM control signal, and directly carrying out maximum power tracking, so that the oscillation process does not exist, and the tracking speed is higher. With the arrival of the big data era, research and optimization algorithms of data mining continuously emerge, photovoltaic MPPT control is performed through a prediction method, and the method has important significance for improving the efficiency of a photovoltaic power generation system.
The photovoltaic cell converts solar energy into electric energy by utilizing a photovoltaic effect, and the output characteristic and the maximum power point of the cell component change along with the change of illumination intensity and temperature. Fig. 1 is a schematic structural diagram of an equivalent circuit of a photovoltaic cell according to an embodiment of the present application. The I-U equation for a photovoltaic cell can be derived from fig. 1:
wherein, Upv、IpvRespectively representing the voltage and the current output by the photovoltaic cell; i isphIs a photo-generated current; i isdRepresenting the diode reverse saturation current; rsRepresents a series equivalent resistance; rpRepresenting the equivalent resistance in parallel, q represents the charge constant, 1.6 × 10-9C, k represents Boltzmann's constant, 1.38 × 10-23J/K; t represents the photovoltaic cell temperature.
Further, according to the above photovoltaic cell equation (formula 1), an I-U, P-U characteristic curve of the photovoltaic cell can be drawn, as shown in fig. 2 to 5. FIG. 2 is a diagram illustrating different I-U characteristics under the change of illumination intensity according to an embodiment of the present application. FIG. 3 is a schematic diagram of different I-U characteristics under the change of temperature conditions according to the embodiment of the present application. FIG. 4 is a diagram illustrating different P-U characteristics under the variation of illumination intensity according to an embodiment of the present application. FIG. 5 is a graph showing different P-U characteristics under varying temperature conditions according to the examples of the present application. As can be seen from fig. 2 to 5, the operation of the photovoltaic cell is affected by the external illumination intensity and temperature conditions to exhibit nonlinear characteristics. When the illumination intensity is reduced, the short-circuit current is reduced obviously, and the open-circuit voltage is not changed greatly, so that the maximum output power of the photovoltaic cell is reduced finally; as the temperature decreases, the output voltage increases, eventually leading to an increase in the maximum output power of the photovoltaic cell. In summary, variations in illumination intensity and temperature can cause changes in the maximum power point of the photovoltaic cell. Thus, the voltage at the maximum power point can be determined by some algorithm, based on the output characteristics of the photovoltaic cells present, providing a reference for system circuit control.
In addition, in the method for realizing the MPPT control of the photovoltaic power generation system by the prediction method in the prior art, a single type model such as a BP neural network model is usually adopted for tracking prediction, and this way enables the efficiency and the response speed of tracking prediction to be affected by the advantages and disadvantages of the single type model, so that the applicability of an actual prediction result is limited.
Disclosure of Invention
In order to solve the technical problem, the invention provides a photovoltaic MPPT control method based on a prediction model, which comprises the following steps: acquiring input data of a prediction model in real time, wherein the input data of the prediction model comprises illumination intensity data and environment temperature data; secondly, determining a maximum power point voltage parameter predicted by each submodel in a preset combined prediction model according to the input data of the prediction model, and further performing combined calculation on the prediction result of each submodel and the corresponding distribution weight according to the distribution weight corresponding to each submodel to obtain a maximum power point combined prediction voltage parameter so as to predict and track the maximum output power point of the current photovoltaic array operation; and thirdly, generating a corresponding pulse control signal according to the current maximum power point combined predicted voltage parameter, inputting the pulse control signal into a boosting drive circuit, and further controlling a grid-connected inverter to complete inversion processing so that the photovoltaic system operates at the maximum power point.
Preferably, the combined predictive model comprises: the device comprises a photovoltaic array BP neural network sub-model, a least square support vector machine sub-model and an extreme learning machine sub-model.
Preferably, in the second step, the variance of the individual prediction result of each submodel is calculated according to training sample data corresponding to each historical prediction control cycle stored in a historical database and the training sample prediction result of each seed model corresponding to each training sample data, where the training sample data includes the prediction model input data acquired in each historical prediction control cycle and the output voltage data of the photovoltaic array acquired at the corresponding time; and according to the variance of the independent prediction result of each submodel, obtaining the distribution weight value corresponding to the independent prediction of each submodel by adopting a variance covariance weight value dynamic distribution method.
Preferably, according to a preset model updating time interval, training each seed model by using the training sample data stored in the current historical database, and further calculating the variance of the individual prediction result of each sub-model by using each trained seed model to adjust the corresponding distribution weight.
Preferably, the predictive model input data further comprises: wind speed data.
Preferably, determining the sample capacity of the training sample data in the current historical database; and calculating the absolute percentage error of the training sample prediction result obtained by predicting each training sample data through each submodel and the corresponding output voltage data, further calculating the average absolute percentage error of each submodel, and obtaining the variance of the independent prediction result of each submodel based on the average absolute percentage error.
In another aspect, a photovoltaic MPPT control system based on a prediction model is provided, which uses the photovoltaic MPPT control method as described above to realize tracking control of a maximum output power point of a photovoltaic array, wherein the system includes: a photovoltaic array; the data acquisition module is used for acquiring input data of a prediction model in real time, wherein the input data of the prediction model comprises illumination intensity data and environment temperature data; the combined prediction module is used for determining a maximum power point voltage parameter predicted by each sub-model in a preset combined prediction model according to the input data of the prediction model, and further performing combined calculation on the prediction result of each sub-model and the corresponding distribution weight according to the distribution weight corresponding to each sub-model to obtain a maximum power point combined prediction voltage parameter so as to predict and track the maximum output power point of the current photovoltaic array operation; the pulse generator is used for generating a corresponding pulse control signal according to the current maximum power point combined predicted voltage parameter and inputting the pulse control signal into the boost driving circuit; the boosting driving circuit is connected with the photovoltaic array and used for controlling the grid-connected inverter to complete inversion processing by using the pulse control signal so that the photovoltaic system operates at the maximum power point; and the grid-connected inverter.
Preferably, the combined predictive model comprises: the device comprises a photovoltaic array BP neural network sub-model, a least square support vector machine sub-model and an extreme learning machine sub-model.
Preferably, the combined prediction module includes a distribution weight value generation sub-module, where the distribution weight value generation sub-module includes: the prediction result variance generating unit is used for calculating the variance of the independent prediction result of each sub-model according to training sample data corresponding to each historical prediction control period stored in a historical database and the training sample prediction result of each seed model corresponding to each training sample data, wherein the training sample data comprises the prediction model input data acquired in each historical prediction control period and the output voltage data of the photovoltaic array acquired at corresponding time; and the weight distribution result generation unit is used for obtaining the distribution weight corresponding to the independent prediction of each sub-model by adopting a variance covariance weight dynamic distribution method according to the variance of the independent prediction result of each sub-model.
Preferably, the prediction result variance generating unit is further configured to train each seed model according to a preset model update time interval by using the training sample data stored in the current historical database, and further calculate, by using each trained seed model, a variance of an individual prediction result of each sub-model to adjust a corresponding allocation weight.
Compared with the prior art, one or more embodiments in the above scheme can have the following advantages or beneficial effects:
the invention provides a photovoltaic MPPT control method and system based on a prediction model. The method and the system perform real-time prediction tracking on the maximum power point of the photovoltaic array by constructing a combined prediction model based on multiple prediction submodels, not only solve the problems of oscillation phenomenon and slow tracking speed in MPPT performed by a traditional method, but also overcome the defect that the actual prediction precision is influenced due to the learning defect of a single submodel, further ensure that the photovoltaic array can accurately and stably operate at the maximum power point, effectively shorten the time for tracking the maximum power point, and have important significance for improving the efficiency of a photovoltaic power generation system. In addition, the combination method enables the prediction result to approach to a prediction submodel with good effect by dynamically distributing the weight, and has stronger adaptability to prediction when the photovoltaic cell is in a complex environment.
While the invention will be described in connection with certain exemplary implementations and methods of use, it will be understood by those skilled in the art that it is not intended to limit the invention to these embodiments. On the contrary, the intent is to cover all alternatives, modifications and equivalents as included within the spirit and scope of the invention as defined by the appended claims.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
fig. 1 is a schematic structural diagram of an equivalent circuit of a photovoltaic cell according to an embodiment of the present application.
FIG. 2 is a diagram illustrating different I-U characteristics under the change of illumination intensity according to an embodiment of the present application.
FIG. 3 is a schematic diagram of different I-U characteristics under the change of temperature conditions according to the embodiment of the present application.
FIG. 4 is a diagram illustrating different P-U characteristics under the variation of illumination intensity according to an embodiment of the present application.
FIG. 5 is a graph showing different P-U characteristics under varying temperature conditions according to the examples of the present application.
Fig. 6 is a step diagram of a photovoltaic MPPT control method based on a prediction model according to an embodiment of the present application.
Fig. 7 is a schematic diagram of a predictive model-based photovoltaic MPPT control method according to an embodiment of the present application.
Fig. 8 is a schematic structural diagram of an overall structure of a photovoltaic MPPT control system based on a prediction model according to an embodiment of the present application.
Fig. 9 is a schematic control principle diagram of a photovoltaic MPPT control system based on a prediction model according to an embodiment of the present application.
Detailed Description
The following detailed description of the embodiments of the present invention will be provided with reference to the drawings and examples, so that how to apply the technical means to solve the technical problems and achieve the technical effects can be fully understood and implemented. It should be noted that, as long as there is no conflict, the embodiments and the features of the embodiments of the present invention may be combined with each other, and the technical solutions formed are within the scope of the present invention.
Additionally, the steps illustrated in the flow charts of the figures may be performed in a computer system such as a set of computer-executable instructions and, although a logical order is illustrated in the flow charts, in some cases, the steps illustrated or described may be performed in an order different than here.
At present, MPPT methods for photovoltaic power generation systems can be divided into two categories: traditional method, predictive method. The constant voltage tracking method and the short-circuit current proportionality coefficient method are used for calculating the maximum power point voltage based on basic approximate rules of the output characteristics of the photovoltaic cell (for example, an approximate linear relation exists between the maximum power point voltage and the open-circuit voltage) so as to achieve the purpose of control. Although the method is simple and easy to implement, the method has large dependence on output characteristics and low efficiency. A disturbance observation method and a conductance incremental method are used for comparing power obtained by calculation according to detected values of output current and voltage of a photovoltaic cell and then carrying out maximum power tracking through voltage loop closed-loop control, and the method also belongs to a self-optimization algorithm. In addition, the prediction method is used for modeling a large amount of historical data by applying a mathematical statistics idea, comprises the steps of predicting by classical mathematics, artificial intelligence and the like, obtaining a voltage value corresponding to the maximum power point through prediction, calculating to obtain a PWM control signal, and directly carrying out maximum power tracking, so that an oscillation process does not exist, and the tracking speed is higher. With the arrival of the big data era, research and optimization algorithms of data mining continuously emerge, photovoltaic MPPT control is performed through a prediction method, and the method has important significance for improving the efficiency of a photovoltaic power generation system.
In the method for realizing MPPT control of the photovoltaic power generation system through a prediction method in the prior art, tracking prediction is usually performed by adopting a single type model such as a BP neural network model, and the like, so that the efficiency and the response speed of the tracking prediction are influenced by the advantages and the disadvantages of the single type model, and the applicability of an actual prediction result is limited.
Therefore, the invention provides a photovoltaic MPPT control method based on a prediction model. According to the method, multiple models capable of predicting the voltage value at the maximum power point of the photovoltaic array are built, the distribution weight values can be dynamically updated, and based on the models, the prediction results of each model and the corresponding distribution weight values are combined and calculated to obtain the voltage reference value at the maximum power point of the photovoltaic array combined with the multiple prediction models. Further, the photovoltaic system is operated at the maximum power point by utilizing components such as a pulse generator, a booster circuit and a grid-connected inverter in the photovoltaic power generation system. The method solves the problems of oscillation phenomenon and low tracking speed of MPPT in the traditional method, and tracks and predicts the maximum power point by combining a combined method of factors and variables influencing photovoltaic output characteristic, thereby not only ensuring that the photovoltaic array can stably run at the maximum power point, but also effectively shortening the time for tracking the maximum power point, and having important significance for improving the efficiency of a photovoltaic power generation system.
Fig. 6 is a step diagram of a photovoltaic MPPT control method based on a prediction model according to an embodiment of the present application. As shown in fig. 6, in step S610, input data of the prediction model during the operation of the photovoltaic power generation system is collected in real time. The input data of the prediction model is used as the data of the real-time input end of the pre-constructed combined prediction model, and at least comprises illumination intensity data and environment temperature data. The combined prediction model is a model capable of predicting the voltage value of the photovoltaic power generation system operating at the maximum power point, and is constructed by a machine learning method.
Because the influence of the temperature on the output characteristic of the photovoltaic module is inevitable in the background art, meanwhile, in order to facilitate the collection of temperature data, in the practical application process, the influence factor of the temperature of the photovoltaic cell is indirectly represented by the ambient temperature. In addition, the influence of the wind speed on the output characteristic of the photovoltaic module can improve the heat dissipation of the photovoltaic module through the air flow, so that the temperature of the photovoltaic cell is reduced, and therefore the wind speed is used as another factor influencing the output characteristic of the photovoltaic array. Further, the predictive model input data further comprises: wind speed data.
In this way, the illumination intensity data, the ambient temperature data and the wind speed data collected in real time are used as the input end nodes of the online (real-time) combined prediction model in the embodiment of the invention, and the final (real-time) prediction result, namely the maximum power point combined prediction voltage parameter, is obtained by using the combined prediction model.
Then, in step S620, according to the currently collected input data of the prediction model, determining a maximum power point voltage parameter predicted by each submodel in the preset combined prediction model, and according to a distribution weight corresponding to each submodel, further performing combined calculation on the (real-time) prediction result of each submodel and the corresponding distribution weight to obtain a maximum power point combined prediction voltage parameter, so as to predict and track the maximum output power point of the current operation of the photovoltaic array. Wherein the combined prediction model comprises at least: the system comprises a photovoltaic array BP neural network sub-model (constructed by optimizing BP through a genetic algorithm), a least square support vector machine sub-model (least square SVM) and an extreme learning machine sub-model (ELM).
It should be noted that, in the sub-model individual prediction, the prediction effect of the genetic algorithm optimized BP is sometimes not lower than that of ELM, but the prediction capability of the BP neural network optimized by the genetic algorithm is doubled for small samples and data with non-uniform sample distribution; the least square SVM can overcome the problems of under-learning and over-learning of BP and ELM, but the training time is relatively long; the prediction fitting effect of the ELM is best, and the algorithm has the characteristic of quick training time, so that the ELM is more agile in the aspect of real-time acquisition and direct prediction. The combination method in the embodiment of the invention utilizes the dynamically adjustable distribution weight to enable the prediction result to approach the prediction submodel with good effect, has stronger adaptability in the complex environment, integrates the advantages of various optimization algorithms, effectively avoids the defects of the prediction submodel and greatly improves the performance of the prediction model.
Fig. 7 is a schematic diagram of a predictive model-based photovoltaic MPPT control method according to an embodiment of the present application. As shown in fig. 7, in step S701, a photovoltaic array BP neural network sub-model in a preset combined prediction model is used, and a maximum power point voltage parameter for the sub-model is obtained from current prediction model input data. Specifically, the illumination intensity data, the environment temperature data and the wind speed data in the input data of the prediction model are respectively input to an input node of a photovoltaic array BP neural network submodel in the combined prediction model, and under the action of the submodel, a corresponding submodel prediction result, namely a maximum power point voltage parameter aiming at the submodel, namely a first prediction result is obtained at an output node of the photovoltaic array BP neural network submodel.
It should be noted that the photovoltaic array BP neural network sub-model is constructed by using historical data of (actual) output voltage data corresponding to the photovoltaic power generation system operating at the maximum power point under different illumination conditions, different environmental temperatures, different wind speeds and corresponding photovoltaic power generation systems as training sample data and optimizing a BP neural network method through a genetic algorithm. The weight and threshold parameters of BP (Back propagation) neural network to each layer are generated by random initialization, the unoptimized values of the parameters reduce the convergence rate when the BP neural network is trained, the predicted value is not a global optimum value, and the genetic algorithm is a search algorithm of global optimization, and the method obtains the optimal individual by optimizing the population one by one. The photovoltaic array BP neural network submodel in the embodiment of the invention is formed by optimizing and constructing the BP neural network through a genetic algorithm, the weight of each layer of the BP neural network and the optimal value of the threshold parameter are obtained through selection, intersection and variation operations, and then the BP neural network prediction is carried out.
The specific steps of the building process of the photovoltaic array BP neural network submodel are as follows:
step 1: and (5) initializing a population. First, a random initial population W ═ W (W)1,W2,…,Wp)TP represents population size; then, for each individual WiDetermining vector w using real number encoding1,w2,…,wlAnd using the chromosome as a chromosome, wherein the individual comprises all the weight values and the threshold values of the BP neural network.
Step 2: and calculating population fitness. Using the weight value and the threshold value of the BP neural network obtained by the chromosome in Step1 as input training BP neural network, obtaining a training prediction output value, and taking the training error square sum as an individual W in the population WiThe fitness of (2).
Step 3: and (6) selecting operation. The roulette method, i.e. the selection strategy based on fitness proportion, is adopted to select the chromosomes in each generation of population. The selection probability of each individual is expressed by the following expression:
wherein, PiDenotes the selection probability of the ith individual, P denotes the population size, fiThe inverse of the fitness value is represented.
Step 4: and (4) performing a crossover operation. Since the individuals are encoded in Step1 using the real number encoding method, accordingly, the real number interleaving is used during the interleaving operationThe method is carried out. Wherein the kth gene and the l gene w are definedlThe interleaving operation at j bits is expressed by the following expressions, respectively:
wkj＝wkj(1-r)+wljr (2)
wlj＝wlj(1-r)+wkjr (3)
wherein r represents a random number between [0, 1 ].
Step 5: and (5) performing mutation operation. The j gene variation result of the i individual is expressed by the following expression:
f(g)＝r2(1-g/Gmax) (5)
wherein, wmax、wminRespectively represent genes wijUpper and lower bounds of the value, G representing the current iteration number, GmaxRepresents the maximum evolution algebra, r represents [0, 1]]Random number of cells r2Representing a random number.
Step 6: after the above operations, the weight and the threshold parameter between the input layer node and the hidden layer node of the optimal BP neural network and the weight and the threshold parameter between the hidden layer node and the output layer node are obtained by satisfying the above constraint conditions. And then, training the parameters as initial parameters of the BP neural network to obtain a photovoltaic array BP neural network submodel, wherein the photovoltaic array BP neural network submodel is used for predicting an optimal first prediction result during online prediction.
In step S702, it is further required to obtain a maximum power point voltage parameter for a sub-model from input data of a current prediction model by using a least squares support vector machine sub-model in a preset combined prediction model. Specifically, the illumination intensity data, the environment temperature data and the wind speed data in the input data of the prediction model are respectively input to an input node of a least square support vector machine submodel in the combined prediction model, and under the action of the submodel, a corresponding submodel prediction result, namely a maximum power point voltage parameter aiming at the submodel, namely a second prediction result is obtained at an output node of the least square support vector machine submodel.
It should be noted that, in the embodiment of the present invention, the least square support vector machine sub-model is constructed by using historical data of (actual) output voltage data corresponding to the photovoltaic power generation system operating at the maximum power point under different illumination conditions, different environmental temperatures, different wind speeds, and by replacing inequality constraints of the support vector machine with equality constraints and converting a quadratic optimization problem in the support vector machine into solving linear equations, where the historical data is used as training sample data (sample set), so that the convergence speed is greatly increased.
The specific steps of the construction process of the least square support vector machine submodel are as follows:
is provided with
Is a sample set, where xiRepresenting an input vector, xi∈Rn，yiRepresenting the corresponding output vector, yi∈RnAnd l represents a sample capacity. Firstly, a least square support vector machine submodel maps a sample to a high-dimensional feature space through a nonlinear function phi, then linear regression is carried out, and the obtained regression function is represented by the following expression:
f(x)＝w·φ(x)+b (6)
in the equation (6), w represents a weight vector, and b represents an offset value. When a least square support vector machine method is adopted to carry out function estimation, epsilon (insensitive loss function) of Vapnik is replaced by a square error loss function, inequality constraints in the support vector machine are replaced by equality constraints, and the optimization problem is expressed by the following expression:
wherein e isiRepresenting sample error terms, C a penalty factor, C>0. Then, a lagrangian function is introduced and expressed by the following expression:
wherein, αiRepresenting a Lagrange multiplier, satisfying the following constraint equation according to optimization conditions:
further, delete eiAnd w, can be:
wherein q isT＝[1,1,…1]，α＝[α1,α2,…αl]TI denotes an identity matrix of order l × l and K denotes a kernel function matrix according to Mercer conditions the kernel function is defined by the following expression:
K(xi,xj)＝φ(xi)Tφ(xj) (11)
where equation (10) is a linear system of equations, α and b can be found, and the regression function of the LSSVM can be further expressed by the following expression:
in the embodiment of the present invention, a Radial Basis Function (RBF) is selected as a kernel function in a process of constructing a least squares support vector machine submodel, and is represented by the following expression:
in formula (13), σ2Representing the kernel width, the parameter reflects the characteristics of the training sample data set, and influences the generalization capability of the submodel. Thus, the construction of the minimum support vector machine submodel is completed through the process, so as to be used inAnd obtaining an optimized second prediction result during line prediction.
Referring to fig. 7, step S703 also needs to use an extreme learning machine sub-model in the preset combined prediction model to obtain a maximum power point voltage parameter for the sub-model from the current prediction model input data. Specifically, the illumination intensity data, the environment temperature data and the wind speed data in the input data of the prediction model are respectively input to the input node of the extreme learning machine submodel in the combined prediction model, and under the action of the submodel, a corresponding submodel prediction result, namely a maximum power point voltage parameter aiming at the submodel, namely a third prediction result is obtained at the output node of the extreme learning machine submodel.
It should be noted that the extreme learning machine sub-model in the embodiment of the present invention is constructed by using historical data of (actual) output voltage data corresponding to the photovoltaic power generation system operating at the maximum power point under different illumination conditions, different environmental temperatures, different wind speeds, and the corresponding photovoltaic power generation system as training sample data (sample set) by a single hidden layer feedforward neural network method. Extreme Learning Machine (ELM) is a new type of single hidden layer feedforward neural network, and has attracted much attention due to its fast training speed. Different from the traditional neural network learning algorithm, the extreme learning machine sub-model has the advantages of high learning speed, better generalization capability and global approximation capability, and solves the problem of slow training of the support vector machine sub-model based on the principle of minimizing the structural risk in the training of large-scale samples.
The specific steps of the construction process of the extreme learning machine submodel are as follows:
is provided with N training samples (x)i，ti) The number of hidden layer nodes is S, wherein xi＝[xi1,xi2,...,xin]T∈Rn，ti＝[ti1,ti2,...,tim]T∈RmAnd S is less than or equal to N, the mathematical model of the extreme learning machine submodel is represented by the following expression:
wherein β ═ β1,...,βS]TRepresenting the output weight vector from hidden node to output node, H (x)j)＝[h1(xj),...,hS(xj)]Represents the nonlinear excitation function and is also the output of the hidden node. Further, hi(xj) Generally expressed by the following expression:
hi(xj)＝G(ai,bi,xj),ai∈Rn,bi∈R (15)
in the present embodiment, G (a)i,bi,xj) A Sigmiod function is selected.
Calculating β connection output weight between hidden layer nodes and output nodes in extreme learning machine submodel＊β is obtained by solving for the minimum approximation error in the squared error＊The value of (c). Wherein the following formula of the minimum approximation error is shown:
in equation (16), T represents an objective matrix of training data, and is expressed by the following expression:
in the formula (17), | · | |, represents the Frobenius norm. Further, the equation (16) is obtained by calculating a least square solution of the minimum norm, that is, by using the expression (18), so as to obtain the above connection output weight, and further construct an extreme learning machine submodel in the embodiment of the present invention, so as to obtain an optimized third prediction result during online prediction. Wherein expression (18) is as follows:
in the formula (18), the reaction mixture,
for a feed-forward network, the smaller the connection weight, the more generalized the generalization ability, β in all least-squares solutions＊With the minimum norm. Therefore, the extreme learning machine submodel can not only reach the minimum approximate error, but also have stronger generalization capability than the traditional gradient descent algorithm, and the generalized mole inverse of H
Is unique, so the solution is also unique.
Next, in step S704, a variance for an individual prediction result of each submodel is calculated according to training sample data corresponding to each historical prediction control period stored in the historical database and a training sample prediction result for each seed model corresponding to each training sample data, and further, an allocation weight corresponding to each submodel individual prediction, that is, a first submodel allocation weight, a second submodel allocation weight, and a third submodel allocation weight, is obtained by using a variance covariance weight dynamic allocation method.
Specifically, firstly, training sample data corresponding to each historical prediction control period stored in a historical database at present are obtained, then, prediction for model updating is carried out on the training sample data through each seed model at present, a corresponding training sample prediction result is obtained, and further, the variance of an individual prediction result of each sub-model is calculated according to the training sample prediction result and output voltage data in the training sample data corresponding to the training sample prediction result. Wherein the training sample data comprises: the control method comprises the steps of collecting illumination intensity data, environment temperature data and wind speed data in each historical prediction control period, and collecting output voltage data corresponding to the maximum power point of the photovoltaic array at corresponding moments. In addition, in each historical prediction period, that is, in each data acquisition period, in addition to the actual output voltage data corresponding to the maximum power point of the photovoltaic array, the output current data corresponding to the maximum power point of the photovoltaic array needs to be acquired, as shown in fig. 9, so as to verify the online prediction result of the photovoltaic array in the offline prediction stage.
It should be noted that, in the embodiment of the present invention, each sub-model for performing model update prediction should be the latest sub-model after performing model update training according to all training sample data stored in the current history database. When the combined prediction model is used for real-time online prediction, output voltage data corresponding to the current maximum power point of the photovoltaic array, which is acquired in real time in each prediction control period (data acquisition period), needs to be integrated with prediction model input data acquired in the same prediction control period to form corresponding training sample data, and the corresponding training sample data is stored in the historical database after the prediction control time (acquisition time) is marked.
Further, when specifically calculating the individual prediction variance of each submodel, it is necessary to first determine the sample capacity of training sample data in the current historical database, then calculate the absolute percentage error between the training sample prediction result obtained after each training sample data is predicted by each submodel (the prediction result here is the prediction result obtained from the training sample data in the historical database that updates the sample capacity after offline training) and the corresponding output voltage data, further calculate the average absolute percentage error of each submodel, and based on this, obtain the variance of the individual prediction result of each submodel by using the following expression (19).
Wherein expression (19) is shown as follows:
in the formula (19), i represents a sub-model number, δiRepresenting the variance of the individual predictions for each submodel, n representing the sample size of training sample data in the current historical database for each submodel, e1、e2…enRespectively representing the prediction result of each training sample corresponding to the corresponding submodel and the absolute value of the output voltage data corresponding to the prediction control timeFor the error in the percentage, the error rate,
and the average absolute percentage error (MAPE) of the output voltage data representing the prediction result of each training sample corresponding to the corresponding sub-model and the corresponding prediction control time.
And then, according to the variance of the result of each submodel individual prediction, obtaining the corresponding distribution weight value of each submodel individual prediction by adopting a variance covariance weight value dynamic distribution method and respectively utilizing an expression (20), an expression (21) and an expression (22). Wherein the expression (20), the expression (21), and the expression (22) are as follows:
w1＝1/[δ1(1/δ1+1/δ2+1/δ3)](20)
w2＝1/[δ2(1/δ1+1/δ2+1/δ3)](21)
w3＝1/[δ3(1/δ1+1/δ2+1/δ3)](22)
wherein, w1、w2、w3Respectively representing the distribution weight values, delta, corresponding to the independent predictions (results) of the photovoltaic array BP neural network submodel, the least square support vector machine submodel and the extreme learning machine submodel1、δ2、δ3And the variances of the individual prediction results of the photovoltaic array BP neural network submodel, the least square support vector machine submodel and the extreme learning machine submodel are respectively represented.
Further, in the embodiment of the present invention, each seed model needs to be trained according to a preset model update time interval by using training sample data stored in the current historical database, and further by using the latest each seed model (which is obtained by completing the training through all training sample data in the current historical database) after the training, the variance of the individual prediction result of each seed model is calculated to adjust the corresponding distribution weight. It should be noted that, in the present invention, the model update time interval is not specifically limited, and those skilled in the art may set the model update time interval according to factors such as the actual CPU load condition and the online prediction accuracy. In one embodiment, considering that the training causes insufficient resources of a CPU and influences normal operation of other work, since the photovoltaic power generation system does not generate power at night, offline model training can be performed at night, the model is updated, and the independent prediction variance and the distribution weight of the corresponding submodel are calculated, so that online prediction is performed by using the updated latest combined prediction model in the power generation period of the photovoltaic power generation system. In addition, if the CPU adopts a controller with higher performance such as dual cores, real-time online update training can be carried out on each sub-model in the power generation period of the photovoltaic power generation system, so that the distributed weight corresponding to each sub-model can be dynamically adjusted in real time and online, and the combined prediction model has better prediction performance.
Finally, referring to fig. 7 again, in step S705, the following expression (23) is further used to perform combined calculation on the first prediction result, the second prediction result, and the third prediction result predicted in real time and the current distribution weight (the first sub-model distribution weight, the second sub-model distribution weight, and the third sub-model distribution weight) corresponding to each sub-model, so as to obtain the maximum power point combined predicted voltage parameter predicted by using the combination method, so as to complete the tracking prediction of the current maximum output power point of the photovoltaic array.
Wherein expression (23) is shown as follows:
p＝w1p1+w2p2+w3p3(23)
in the formula (23), p represents a maximum power point combined predicted voltage parameter (combined predicted result), and p1、p2、p3Respectively showing the first prediction result, the second prediction result and the third prediction result.
It can be seen from the above expressions (19) to (23) that the larger the variance of the calculated individual prediction result of each submodel is, that is, the larger the absolute error between the prediction result and the actual output voltage of the photovoltaic array is, the smaller the weight value of the corresponding submodel is, so that the final prediction result approaches to the predictor model with better effect. Therefore, the weight dynamic allocation and adjustment method has stronger adaptability to the prediction result of the photovoltaic array under the complex environment condition.
Further, referring to fig. 6 again, in step S630, a corresponding pulse control signal is generated according to the current maximum power point combination predicted voltage parameter, and the pulse control signal is input into the boost driving circuit, and the grid-connected inverter is further controlled to complete the inversion process, so that the photovoltaic system operates at the maximum power point, and the maximum output power of the photovoltaic system is provided to the grid side.
Specifically, according to the input data of the prediction model collected in real time, the maximum power point combination prediction voltage parameter output by the combination prediction model is utilized, and the pulse generator 40 (i.e., the PWM control module) in the photovoltaic power generation system calculates the duty ratio based on the parameter, so as to obtain the pulse control signal matched with the maximum power point combination prediction voltage parameter. Further, the current pulse control signal is input to an IGBT trigger terminal in the boost driving circuit 50 connected to the output terminal of the photovoltaic array 10, and the on-off state of the IGBT is controlled, so that the boost driving circuit 50 is used to control the grid-connected inverter 60 to complete the corresponding inversion process, and the entire photovoltaic power generation system is operated at the maximum power point, thereby tracking the maximum power, and further providing the maximum operating power to the grid side (not shown). Therefore, the aim of maximum power tracking is achieved, the process of the current prediction control is finished, and the prediction control of the MTTP is continued according to the same method in the next prediction period.
In addition, the invention also provides a photovoltaic MPPT control system (also called a photovoltaic power generation system) based on the prediction model, and the system realizes the tracking control of the maximum output power point of the photovoltaic array by using the photovoltaic MPPT control method. Fig. 8 is a schematic structural diagram of an overall structure of a photovoltaic MPPT control system based on a prediction model according to an embodiment of the present application. As shown in fig. 8, the system includes: the photovoltaic array 10, the data acquisition module 20, the combined prediction module 30, the pulse generator 40, the boost driving circuit 50 and the grid-connected inverter 60. Fig. 9 is a schematic control principle diagram of a photovoltaic MPPT control system based on a prediction model according to an embodiment of the present application. The modules and devices in the system are described below with reference to fig. 8 and 9.
The photovoltaic array 10 is used for outputting output current data and output voltage data corresponding to the maximum power point in real time under the influence of the current ambient temperature, the illumination intensity and the wind speed.
The data acquisition module 20 includes devices such as a temperature sensor, an ultraviolet sensor, a wind speed measuring instrument, a current sensor, and a voltage sensor, and is respectively configured to acquire, in real time, illumination intensity data, ambient temperature data, and wind speed data in the input data of the prediction model, and output voltage data and output current data corresponding to the maximum power point of the photovoltaic array, which are acquired in real time.
The combined prediction module 30 is connected to the data acquisition module 20, and is configured to determine a maximum power point voltage parameter predicted by each sub-model in a preset combined prediction model according to the input data of the prediction model, and further perform combined calculation on the prediction result of each sub-model and a corresponding distribution weight according to the distribution weight corresponding to each sub-model to obtain a maximum power point combined prediction voltage parameter, so as to predict and track a maximum output power point of the current photovoltaic array operation. Wherein, the combined prediction model comprises: the device comprises a photovoltaic array BP neural network sub-model, a least square support vector machine sub-model and an extreme learning machine sub-model. Further, the combination prediction module 30 includes: a first prediction submodule 31, a second prediction submodule 32, a third prediction submodule 33, an allocation weight value generation submodule 34, and a combined prediction calculation result generation submodule 35.
Specifically, the first prediction sub-module 31 is implemented according to the method described in step S701, and is configured to obtain, by using a photovoltaic array BP neural network sub-model in a preset combined prediction model, a maximum power point voltage parameter for the sub-model, that is, a first prediction result, from data input by the current prediction model. The second prediction sub-module 32 is implemented according to the method described in step S702, and is configured to obtain, from the current prediction model input data, a maximum power point voltage parameter, i.e., a second prediction result, for a sub-model of a least squares support vector machine in a preset combined prediction model. The third prediction sub-module 33 is implemented according to the method described in the above step S703, and is configured to obtain, from the current prediction model input data, a maximum power point voltage parameter for a preset limit learning machine sub-model in the combined prediction model, that is, a third prediction result.
In addition, the assignment weight value generation sub-module 34 is implemented according to the method described in step S704, and includes a prediction result variance generation unit 341 and a weight value assignment result generation unit 342. The prediction result variance generating unit 341 is configured to calculate a variance of an individual prediction result for each sub-model according to the training sample data corresponding to each historical prediction control period stored in the historical database and the training sample prediction result for each seed model corresponding to each training sample data. The weight distribution result generating unit 342 is configured to obtain the distribution weight corresponding to each submodel for individual prediction by using a variance covariance weight dynamic distribution method according to the variance of the individual prediction result of each submodel.
In addition, the combined prediction module 30 further includes a combined prediction calculation result generation sub-module 35, where the sub-module 35 is implemented according to the method described in the step S705, and is configured to perform combined calculation on the first prediction result, the second prediction result, and the third prediction result predicted in real time and the current allocation weight corresponding to each sub-model by using the expression (23), so as to obtain a maximum power point combined prediction voltage parameter predicted by using a combination method, so as to complete tracking prediction of the current maximum output power point of the photovoltaic array.
The pulse generator 40 is connected to the combined prediction module 30. The pulse generator 40 is configured to combine the predicted voltage parameters according to the predicted maximum power point in the current prediction period, perform duty ratio calculation on the parameters, generate corresponding pulse control signals, and input the pulse control signals into the boost driving circuit 50.
Referring to fig. 9, a boost driver circuit 50 is connected across the output terminals of the photovoltaic array 10. Further, a trigger input port of the power driver in the boost driving circuit 50 is connected to an output terminal of the pulse generator 30. The boost driving circuit 50 is configured to receive the pulse control signal and control the grid-connected inverter 60 to perform corresponding inversion processing, so that the photovoltaic system operates at a maximum power point, thereby tracking the maximum power point. The boost driving circuit 50 receives and uses the pulse control signal to control the on/off of the power driving device in the circuit 50, thereby controlling the grid-connected inverter 60. In the embodiment of the present invention, the Boost driving circuit 50 adopts a Boost circuit structure, and the power driving device adopts an IGBT device, that is, the purpose of maximum power tracking is achieved by controlling the on/off of the IGBT.
Further, the grid-connected inverter 60 is connected across the output end of the boost driving circuit 50, and is configured to perform an inversion process on the actual voltage corresponding to the current maximum power point combined predicted voltage parameter under the control of the boost driving circuit 50, and then output the actual voltage to the grid side (not shown).
Therefore, the purpose of predicting and tracking the maximum power point is achieved through the structure of the photovoltaic MPPT control system in the embodiment of the invention.
The invention provides a photovoltaic MPPT control method and system based on a prediction model. The method and the system perform real-time prediction tracking on the maximum power point of the photovoltaic array by constructing a combined prediction model based on multiple prediction submodels, not only solve the problems of oscillation phenomenon and slow tracking speed in MPPT performed by a traditional method, but also overcome the defect that the actual prediction precision is influenced due to the learning defect of a single submodel, further ensure that the photovoltaic array can accurately and stably operate at the maximum power point, effectively shorten the time for tracking the maximum power point, and have important significance for improving the efficiency of a photovoltaic power generation system. In addition, the combination method enables the prediction result to approach to a prediction submodel with good effect by dynamically distributing the weight, and has stronger adaptability to prediction when the photovoltaic cell is in a complex environment.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A photovoltaic MPPT control method based on a prediction model comprises the following steps:
acquiring input data of a prediction model in real time, wherein the input data of the prediction model comprises illumination intensity data and environment temperature data;
secondly, determining a maximum power point voltage parameter predicted by each submodel in a preset combined prediction model according to the input data of the prediction model, and further performing combined calculation on the prediction result of each submodel and the corresponding distribution weight according to the distribution weight corresponding to each submodel to obtain a maximum power point combined prediction voltage parameter so as to predict and track the maximum output power point of the current photovoltaic array operation;
and thirdly, generating a corresponding pulse control signal according to the current maximum power point combined predicted voltage parameter, inputting the pulse control signal into a boosting drive circuit, and further controlling a grid-connected inverter to complete inversion processing so that the photovoltaic system operates at the maximum power point.
2. The photovoltaic MPPT control method of claim 1, wherein the combined predictive model includes: the device comprises a photovoltaic array BP neural network sub-model, a least square support vector machine sub-model and an extreme learning machine sub-model.
3. The photovoltaic MPPT control method of claim 1 or 2, characterized in that in the second step,
calculating the variance of the independent prediction result of each sub-model according to training sample data corresponding to each historical prediction control period stored in a historical database and the training sample prediction result corresponding to each training sample data and aiming at each sub-model, wherein the training sample data comprises the prediction model input data acquired in each historical prediction control period and the output voltage data of the photovoltaic array acquired at corresponding time;
and according to the variance of the independent prediction result of each submodel, obtaining the distribution weight value corresponding to the independent prediction of each submodel by adopting a variance covariance weight value dynamic distribution method.
4. The photovoltaic MPPT control method of claim 3, wherein each seed model is trained using the training sample data stored in a current historical database according to a preset model update time interval, and further using each trained seed model to calculate a variance of an individual prediction result of each sub model to adjust a corresponding distribution weight.
5. The photovoltaic MPPT control method of any one of claims 1-4, wherein the predictive model input data further comprises: wind speed data.
6. The photovoltaic MPPT control method of claim 3 or 4,
determining the sample capacity of the training sample data in the current historical database;
and calculating the absolute percentage error of the training sample prediction result obtained by predicting each training sample data through each submodel and the corresponding output voltage data, further calculating the average absolute percentage error of each submodel, and obtaining the variance of the independent prediction result of each submodel based on the average absolute percentage error.
7. A photovoltaic MPPT control system based on a prediction model, which realizes tracking control of a photovoltaic array maximum output power point by using the photovoltaic MPPT control method according to any one of claims 1-6, wherein the system comprises:
a photovoltaic array;
the data acquisition module is used for acquiring input data of a prediction model in real time, wherein the input data of the prediction model comprises illumination intensity data and environment temperature data;
the combined prediction module is used for determining a maximum power point voltage parameter predicted by each sub-model in a preset combined prediction model according to the input data of the prediction model, and further performing combined calculation on the prediction result of each sub-model and the corresponding distribution weight according to the distribution weight corresponding to each sub-model to obtain a maximum power point combined prediction voltage parameter so as to predict and track the maximum output power point of the current photovoltaic array operation;
the pulse generator is used for generating a corresponding pulse control signal according to the current maximum power point combined predicted voltage parameter and inputting the pulse control signal into the boost driving circuit;
the boosting driving circuit is connected with the photovoltaic array and used for controlling the grid-connected inverter to complete inversion processing by using the pulse control signal so that the photovoltaic system operates at the maximum power point; and
the grid-connected inverter is provided.
8. The photovoltaic MPPT control system of claim 7, wherein the combined predictive model includes: the device comprises a photovoltaic array BP neural network sub-model, a least square support vector machine sub-model and an extreme learning machine sub-model.
9. The photovoltaic MPPT control system of claim 7 or 8, wherein the combined prediction module includes an assignment weight generation submodule, wherein the assignment weight generation submodule includes:
the prediction result variance generating unit is used for calculating the variance of the independent prediction result of each sub-model according to training sample data corresponding to each historical prediction control period stored in a historical database and the training sample prediction result of each seed model corresponding to each training sample data, wherein the training sample data comprises the prediction model input data acquired in each historical prediction control period and the output voltage data of the photovoltaic array acquired at corresponding time;
and the weight distribution result generation unit is used for obtaining the distribution weight corresponding to the independent prediction of each sub-model by adopting a variance covariance weight dynamic distribution method according to the variance of the independent prediction result of each sub-model.
10. The MPPT control system of claim 9, wherein the variance of the prediction result generation unit is further configured to train each seed model according to a preset model update time interval by using the training sample data stored in the current historical database, and further calculate the variance of the individual prediction result of each sub model by using each trained seed model to adjust the corresponding distribution weight.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596118.XA CN111367349A (en) | 2018-12-26 | 2018-12-26 | Photovoltaic MPPT control method and system based on prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811596118.XA CN111367349A (en) | 2018-12-26 | 2018-12-26 | Photovoltaic MPPT control method and system based on prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111367349A true CN111367349A (en) | 2020-07-03 |
Family
ID=71210208
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811596118.XA Pending CN111367349A (en) | 2018-12-26 | 2018-12-26 | Photovoltaic MPPT control method and system based on prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111367349A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112564098A (en) * | 2020-12-02 | 2021-03-26 | 国网浙江省电力有限公司电力科学研究院 | High-proportion photovoltaic power distribution network voltage prediction method based on time convolution neural network |
| CN113050746A (en) * | 2021-03-24 | 2021-06-29 | 温州大学 | Maximum power tracking method of photovoltaic power generation system based on memory enhancement |
| CN113177362A (en) * | 2021-05-20 | 2021-07-27 | 上海华谊信息技术有限公司 | Furnace temperature prediction method and device based on furnace temperature soft measurement model |
| CN114465577A (en) * | 2022-04-12 | 2022-05-10 | 华东交通大学 | Photovoltaic photo-thermal-based solar energy utilization device and optimal control method |
| CN114510111A (en) * | 2021-12-29 | 2022-05-17 | 北京华能新锐控制技术有限公司 | Global MPPT control method and device for partial sun-shading photovoltaic array |
| NL2028909B1 (en) * | 2021-08-03 | 2022-07-06 | Univ Central South | Method for maximum power point tracking |
| CN114764262A (en) * | 2021-01-11 | 2022-07-19 | 领鞅科技(杭州)有限公司 | Method for predicting and controlling power generation power of solar power station |
| CN115951755A (en) * | 2023-02-06 | 2023-04-11 | 广芯微电子（广州）股份有限公司 | Photovoltaic maximum power point tracking method and device based on PSO-LSTM |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102385724A (en) * | 2010-08-27 | 2012-03-21 | 上海财经大学 | Spare part assembling demand forecasting information processing method applied to inventory management |
| CN103823504A (en) * | 2014-03-20 | 2014-05-28 | 常州工学院 | Maximum power tracing control method based on least squares support vector machine |
| CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
| CN105184678A (en) * | 2015-09-18 | 2015-12-23 | 齐齐哈尔大学 | Method for constructing photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms |
| CN106803135A (en) * | 2017-01-17 | 2017-06-06 | 广东工业大学 | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power |
| CN107169683A (en) * | 2017-06-30 | 2017-09-15 | 上海电力学院 | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient |
| CN108233417A (en) * | 2018-01-22 | 2018-06-29 | 华北电力大学（保定） | A kind of regenerative resource of more prediction model cascade compensation parallel coupleds, which is contributed, predicts computational methods |
-
2018
- 2018-12-26 CN CN201811596118.XA patent/CN111367349A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102385724A (en) * | 2010-08-27 | 2012-03-21 | 上海财经大学 | Spare part assembling demand forecasting information processing method applied to inventory management |
| CN103823504A (en) * | 2014-03-20 | 2014-05-28 | 常州工学院 | Maximum power tracing control method based on least squares support vector machine |
| CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
| CN105184678A (en) * | 2015-09-18 | 2015-12-23 | 齐齐哈尔大学 | Method for constructing photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms |
| CN106803135A (en) * | 2017-01-17 | 2017-06-06 | 广东工业大学 | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power |
| CN107169683A (en) * | 2017-06-30 | 2017-09-15 | 上海电力学院 | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient |
| CN108233417A (en) * | 2018-01-22 | 2018-06-29 | 华北电力大学（保定） | A kind of regenerative resource of more prediction model cascade compensation parallel coupleds, which is contributed, predicts computational methods |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112564098A (en) * | 2020-12-02 | 2021-03-26 | 国网浙江省电力有限公司电力科学研究院 | High-proportion photovoltaic power distribution network voltage prediction method based on time convolution neural network |
| CN114764262A (en) * | 2021-01-11 | 2022-07-19 | 领鞅科技(杭州)有限公司 | Method for predicting and controlling power generation power of solar power station |
| CN114764262B (en) * | 2021-01-11 | 2023-08-15 | 领鞅科技(杭州)有限公司 | Solar power station power generation power prediction and control method |
| CN113050746A (en) * | 2021-03-24 | 2021-06-29 | 温州大学 | Maximum power tracking method of photovoltaic power generation system based on memory enhancement |
| CN113177362A (en) * | 2021-05-20 | 2021-07-27 | 上海华谊信息技术有限公司 | Furnace temperature prediction method and device based on furnace temperature soft measurement model |
| NL2028909B1 (en) * | 2021-08-03 | 2022-07-06 | Univ Central South | Method for maximum power point tracking |
| CN114510111A (en) * | 2021-12-29 | 2022-05-17 | 北京华能新锐控制技术有限公司 | Global MPPT control method and device for partial sun-shading photovoltaic array |
| CN114510111B (en) * | 2021-12-29 | 2023-09-12 | 北京华能新锐控制技术有限公司 | Global MPPT control method and device for partial shading photovoltaic array |
| CN114465577A (en) * | 2022-04-12 | 2022-05-10 | 华东交通大学 | Photovoltaic photo-thermal-based solar energy utilization device and optimal control method |
| CN115951755A (en) * | 2023-02-06 | 2023-04-11 | 广芯微电子（广州）股份有限公司 | Photovoltaic maximum power point tracking method and device based on PSO-LSTM |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111367349A (en) | Photovoltaic MPPT control method and system based on prediction model | |
| Yap et al. | Artificial intelligence based MPPT techniques for solar power system: A review | |
| Gao et al. | Day-ahead power forecasting in a large-scale photovoltaic plant based on weather classification using LSTM | |
| Karabacak et al. | Artificial neural networks for controlling wind–PV power systems: A review | |
| Karlis et al. | A novel maximum power point tracking method for PV systems using fuzzy cognitive networks (FCN) | |
| CN110738344B (en) | Distributed reactive power optimization method and device for load prediction of power system | |
| Mellit et al. | Application of neural networks and genetic algorithms for sizing of photovoltaic systems | |
| KR100886891B1 (en) | Control system of solar cell generation using genetic algorithm and neuro fuzzy controller | |
| CN111222674A (en) | Short-term photovoltaic power generation capacity prediction method based on long-term and short-term memory neural network | |
| CN112491094B (en) | Hybrid-driven micro-grid energy management method, system and device | |
| Aldobhani et al. | Maximum power point tracking of PV system using ANFIS prediction and fuzzy logic tracking | |
| CN104217113B (en) | Independent wind-light storage system reliability estimation method based on energy storage probabilistic model | |
| CN108092321B (en) | Active power and reactive power coordinated control method considering uncertainty for active power distribution network | |
| WO2019006995A1 (en) | Intelligent prediction system for power battery soc of electric vehicle | |
| CN106779175B (en) | Time-interval-considered nonlinear photovoltaic output prediction method | |
| Fatima et al. | Intelligent approach-based maximum power point tracking for renewable energy system: a review | |
| Chatziagorakis et al. | Enhancement of hybrid renewable energy systems control with neural networks applied to weather forecasting: the case of Olvio | |
| CN116862050A (en) | Time sequence network-based daily prediction method, system, storage medium and equipment for carbon emission factors | |
| CN112926653B (en) | Photovoltaic power short-term prediction method and system based on small-scale weather type correction | |
| CN116542882B (en) | Photovoltaic power generation smoothing method, system and storage medium | |
| Wongsathan et al. | Optimal hybrid neuro-fuzzy based controller using MOGA for photovoltaic (PV) battery charging system | |
| CN117634737A (en) | Photovoltaic power generation prediction model construction method, photovoltaic energy supply regulation and control method and device | |
| Zerouali et al. | A robust fuzzy logic PI controller for solar system battery charging | |
| Karatepe et al. | Fuzzy wavelet network identification of optimum operating point of non-crystalline silicon solar cells | |
| CN109217357A (en) | A kind of grid-connected photovoltaic system MPPT method based on Markov model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200703 |
|
| RJ01 | Rejection of invention patent application after publication |