CN115411749A - Photovoltaic power smoothing method considering power prediction - Google Patents
Photovoltaic power smoothing method considering power prediction Download PDFInfo
- Publication number
- CN115411749A CN115411749A CN202211352600.5A CN202211352600A CN115411749A CN 115411749 A CN115411749 A CN 115411749A CN 202211352600 A CN202211352600 A CN 202211352600A CN 115411749 A CN115411749 A CN 115411749A
- Authority
- CN
- China
- Prior art keywords
- power
- smoothing
- moment
- energy storage
- value
- 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
- 238000009499 grossing Methods 0.000 title claims abstract description 99
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000004146 energy storage Methods 0.000 claims abstract description 63
- 238000012549 training Methods 0.000 claims abstract description 10
- 238000012937 correction Methods 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 125000004432 carbon atom Chemical group C* 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 21
- 238000007599 discharging Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 7
- 238000010248 power generation Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 230000009466 transformation Effects 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000004140 cleaning Methods 0.000 description 3
- 150000001875 compounds Chemical class 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003749 cleanliness Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/24—Arrangements for preventing or reducing oscillations of power in networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/28—Arrangements for balancing of the load in a network by storage of energy
- H02J3/32—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/381—Dispersed generators
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/22—The renewable source being solar energy
- H02J2300/24—The renewable source being solar energy of photovoltaic origin
-
- 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
Abstract
The invention provides a photovoltaic power smoothing method considering power prediction, which comprises the following steps: acquiring photovoltaic power historical data, and training a prediction model; the method comprises the steps of collecting photovoltaic power data of an inverter, adopting a smoothing algorithm to the collected data at the moment to obtain a power smooth value at the moment and an energy storage SOC value at the moment, predicting the photovoltaic power at the next moment based on a trained prediction model, similarly adopting the smoothing algorithm to obtain the power smooth value at the next moment and the energy storage SOC value at the next moment, judging whether the energy storage SOC value at the next moment is in a limited range, if so, outputting the power smooth value at the moment and the energy storage SOC value at the moment, and if not, correcting the power smooth value at the moment. According to the method, the power prediction process and the power smoothing process are combined, and compared with the existing photovoltaic power smoothing strategy, the time of the energy storage battery under the limit SOC condition can be shortened, and the working life of energy storage is maintained.
Description
Technical Field
The invention mainly relates to the technical field related to data processing in a photovoltaic power generation system, in particular to a photovoltaic power smoothing method considering power prediction.
Background
Photovoltaic power generation has cleanliness and flexibility, and compared with other forms of new energy, the resource is abundant, more stable, has consequently obtained rapid development. As the permeability of photovoltaic power generation is continuously increased, the large fluctuation of photovoltaic power will have a large influence on the stability of the power system.
The photovoltaic power smoothing algorithm commonly used at present comprises a low-pass filter algorithm, a moving average method, a slope control method and the like. The low-pass filtering algorithm and the moving average algorithm have large phase lag, and the slope control method has the problem of control precision. The TD (steepest tracking differentiator) algorithm can greatly reduce the phase lag phenomenon on the basis of better smoothing effect, and is applied to photovoltaic power smoothing. The photovoltaic power output is time sequence data, the photovoltaic power output at the previous moment can influence the photovoltaic power smoothing condition at the later moment, however, the combination of photovoltaic power prediction and power smoothing is not considered in the algorithm, so that the time of the energy storage battery under the condition of limit SOC is longer, and the maintenance of the working life of energy storage is not facilitated. Based on this, the photovoltaic power prediction result is used for guiding the power smoothing process, the method is suitable for the smoothing process under the conditions of large power fluctuation and poor energy storage initial electric quantity, the smoothing value after TD is further corrected by adopting a strategy, and finally the service life of the energy storage battery is prolonged.
Disclosure of Invention
In order to solve the defects of the prior art, the invention provides a photovoltaic power smoothing method considering power prediction from the practical application by combining the prior art, and compared with the existing photovoltaic power smoothing strategy, the method can reduce the time of the energy storage battery under the condition of the limit SOC and is beneficial to maintaining the working life of energy storage by combining the power prediction with the power smoothing process.
In order to realize the purpose, the technical scheme of the invention is as follows:
a photovoltaic power smoothing method taking into account power prediction, comprising the steps of:
s1: acquiring photovoltaic power historical data, and training a prediction model;
s2: acquiring photovoltaic power data of the inverter, and obtaining a power smooth value at the moment and an energy storage SOC value at the moment after smoothing by adopting a smoothing algorithm on the acquired data at the moment;
s3: predicting the photovoltaic power at the next moment based on the trained prediction model, and obtaining a power smooth value at the next moment and an energy storage SOC value at the next moment after smoothing by adopting a smoothing algorithm;
s4: judging whether the energy storage SOC value at the next moment is in a limited range,
if so, outputting the power smooth value and the energy storage SOC value at the moment,
and if not, correcting the power smooth value at the moment.
Further, the photovoltaic power historical data is divided into a plurality of subsequences by the prediction model through multi-order wavelet transformation, the subsequences comprise approximate components and detail components, different CNN-LSTM neural networks are used for predicting different subsequences, and the final prediction result is subjected to inverse normalization and inverse wavelet transformation to obtain a prediction result related to power.
Further, in step S4, when the energy storage SOC value at the next moment is smoothed according to the algorithm and reaches the upper limit, it is determined that the overcharge phenomenon may be caused at the future moment, and the energy storage battery is discharged at the current moment.
Further, in step S4, the energy storage SOC value at the next time is defined as 90% at the upper line and 10% at the lower line.
Further, in step S4, when the power smooth value at this time is corrected, the smooth correction reference value is as follows:
is provided withJTo representtAt points in the time-of-day windowPower value,Representing the rated capacity of the photovoltaic array, SOC (t + 1) representing the energy storage SOC value at the next moment, and the upper limit and the lower limit of the defined SOC (t + 1) are respectively 90 percent and 10 percent,
SOC (t + 1) > 90%:
When the temperature is higher than the set temperatureAnd is andwhen the reference value is smoothly corrected to be;
When SOC (t + 1) < 10%:
when Max (J)-10%P u >Min(J)+10%P u The smooth correction reference value is Min (J)+10%P u ;
When Min: (J) <Max(J)-10%P u <Min(J) +10%P u And Max: (J)-10%P u <P g (t) When the smoothed modified reference value is Max: (J)-10%P u ;
When Min: (J) <Max(J)-10%P u <Min(J) +10%P u And Max (Max)J)-10%P u >P g (t) The smooth corrected reference value is Min: (J);
When Max (J)-10%P u < Min(J) And Min: (J)<P g (t) The smooth correction reference value is Min (J);
When Max (J)-10%P u < Min(J) And Min: (J)>P g (t) When the smoothed modified reference value is Max: (J) -10%P u 。
Further, when the power is smooth, a power fluctuation evaluation index of 10s level in a short time scale is adopted, and the evaluation index is as follows:
in the above formula, the first and second carbon atoms are,for the photovoltaic power after the energy storage smoothing,the output power of the photovoltaic array sampled at time i.
Further, the sampling time interval is set to 10s, and the communication protocol adopts a Modbus communication protocol.
Furthermore, the smoothing algorithm adopts a steepest tracking differentiator, and the steepest tracking differentiator is the input of a traditional tracking differentiatoruSet to a particular non-linear formfhan,The expression is as follows:
wherein the content of the first and second substances,is composed offhanThe difference between the output and the input signal,is composed offhanAn estimate of the derivative of the input signal is made, fhanin which there are two parameter tracking factors to be adjustedFilter factorThe input signal passes throughfhanThereafter, a noise-free input tracking signal and an estimate of the derivative of the input signal may be obtained.
The invention has the beneficial effects that:
1. the method combines the photovoltaic power prediction with the photovoltaic power smoothing algorithm, provides a strategy, smoothes the photovoltaic power under the premise of considering the power prediction, can reduce the time of the energy storage battery in the limit state of charge, and is beneficial to maintaining the service life of the energy storage battery.
2. The specific power smooth value correction strategy provided by the invention fully considers the smooth value as far as possible within the upper and lower line ranges and the corrected smooth value as far as possible, and the fluctuation rate, and the two problems of the same fluctuation rate as that of the TD algorithm are ensured as far as possible, so that the correction strategy can reduce the time of the stored energy in the limit charge state, and is easy to implement in engineering.
3. The method adopts discrete wavelet transformation combined with the CNN-LSTM neural network to predict the photovoltaic power generation power, divides the power generation prediction into different frequency band data to predict, and improves the prediction precision.
4. Compared with the traditional minute-level power fluctuation index, the invention provides a 10 s-level power fluctuation index, which is more favorable for finding the hidden power of the photovoltaic power in a certain sense, thereby being more favorable for smoothing the photovoltaic power.
5. The invention carries out smoothing based on TD algorithm, and has better tracking effect compared with the traditional low-pass filter and moving average algorithm due to the nonlinear structure.
Drawings
FIG. 1 is a flow chart of a smoothing strategy;
FIG. 2 is a DWT-CNN-LSTM photovoltaic power model prediction flow;
FIG. 3 is a specific example of a photovoltaic power smoothing strategy that takes power prediction into account;
FIG. 4 is a graph of raw photovoltaic power;
fig. 5 shows the smoothing effect (SOC (t + 1) > 90%) after applying the proposed strategy;
FIG. 6 shows the SOC change (SOC (t + 1) > 90%) after applying the proposed strategy;
fig. 7 is the smoothing effect after applying the proposed strategy (SOC (t + 1) < 10%);
fig. 8 shows the change in SOC after applying the proposed strategy (SOC (t + 1) < 10%).
Detailed Description
The invention is further described with reference to the accompanying drawings and the specific embodiments. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and these equivalents also fall within the scope of the present application.
The embodiment provides a photovoltaic power smoothing method considering power prediction, and based on the industrial situation, the whole-day power smoothing will cause large variation of the SOC state of the energy storage battery. For example, the photovoltaic output power fluctuation is large in cloudy days, the energy storage battery is adopted for smoothing, so that the energy storage charging and discharging power is large in the whole day, and when the initial SOC of the energy storage is too large or too small, the maximum charge state can be easily reached in the smoothing process. In order to prevent the energy storage battery from being excessively charged and discharged, the SOC is generally limited in the power smoothing process, and in this embodiment, the limit of the SOC is specifically set to [10%,90% ]. The charging and discharging strategy of the energy storage in the traditional process is as follows:
in the formula (I), the compound is shown in the specification,represents the discharge power of the energy storage battery,representing the charging power of the energy storage battery. According to the formula, in the actual power smoothing process, if the initial SOC condition is poor, the adoption of the strategy can lead the energy storage to work in the limit state of charge for a long time, and the service life of the energy storage battery is greatly reduced, so that the photovoltaic power smoothing strategy combined with power prediction has a great significance.
Because the photovoltaic power is a time sequence data, under the condition that the energy storage electric quantity is certain, the smoothing result of the photovoltaic power at the previous moment can affect the power smoothing process at the later moment, the embodiment provides a photovoltaic power smoothing strategy by considering the photovoltaic power prediction result on the basis of the photovoltaic power smoothing algorithm, and corrects the power smoothing reference value when the energy storage battery electric quantity is at the limit according to the future prediction result, so as to reserve space for charging and discharging at the next moment. As shown in fig. 1, the photovoltaic power smoothing method capable of considering power prediction provided in this embodiment is applied to photovoltaic power smoothing operation, and the main design idea is to calculate the battery energy storage SOC condition after charging and discharging through power smoothing operation at this time; and predicting the power value at the next t +1 moment through a trained model, calculating the SOC condition by adopting the same power smoothing operation, judging the SOC after charging and discharging at the t +1 moment, detecting whether the SOC is in a limit SOC condition, and correcting the smoothing result at the t moment according to a rule if the SOC is in the limit SOC condition.
The smoothing method provided by the implementation adopts the following specific steps in the practical application of photovoltaic power generation:
the method comprises the following steps: collecting information in a field inverter, arranging RS485 wiring, communicating through an RS485 special line at the bottom layer, communicating by WiFi in the upper computer, collecting field data in real time, uploading the field data to a Mysql database of the upper computer for storage, and marking a time tag at the same time, in order to meet the designed short time scale fitness function, in the embodiment, the sampling time interval is set to be 10s, the communication protocol adopts a Modbus communication protocol,
in order to prevent communication interruption, a breakpoint reconnection method is adopted, reconnection is automatically carried out at the moment of interruption, and data acquisition and the whole smoothing process are guaranteed.
Step two: displaying real-time acquired data, regularly cleaning historical data in a database, supplementing and deleting repeated data to the acquired data, mainly sorting the acquired data by taking a time tag as a basis in the cleaning process, and checking and cleaning the data in the database at a time interval of 3 days;
step three: smoothing the acquired data by adopting a TD algorithm, and selecting parameters asStep sizeFilter factorCalculating the power smooth value X at this moment t And the SOC of the energy storage battery after charging and discharging t (ii) a condition;
step four: the on-site collected data is predicted in real time through a trained prediction model, smoothing is carried out by adopting a TD algorithm, and a power smoothing value X at the next moment is calculated t+1 And the SOC of the energy storage battery after charging and discharging t+1 A condition;
step five: observing whether the SOC in the prediction period exceeds the limit, and if so, correcting the power smooth condition at the moment to prepare for charging and discharging in the next time period;
step six: and forming a whole set of photovoltaic power smoothing flow, and displaying the acquired data information, the battery information, the prediction information and the smoothing information to the platform in real time. The method can be realized by adopting a Python Tkinter library development C/S software platform architecture.
In the prediction model of the embodiment, the bottom layer collects inverter data and battery data and sends the inverter data and the battery data to the upper computer, the upper computers are connected through TCP/IP, the upper computer can acquire system data in real time, and the prediction model can be used for training the photovoltaic power prediction model under the condition of acquiring a large amount of historical data.
The overall architecture of the power prediction model of the present embodiment is shown in fig. 2. The CNN-LSTM neural network is adopted in the prediction process, and in order to achieve a good prediction effect, data are preprocessed before a prediction model is trained, and discrete wavelet decomposition is adopted for processing. Specifically, three-order wavelet transformation is adopted to convert original photovoltaic power dataDivided into a plurality of subsequences, including approximate componentsAnd detail componentAfter the third-order wavelet transform, the original signal can be represented by adopting a decomposition subsequence thereof:
and adopting different CNN-LSTM neural networks for prediction for different subsequences, and performing inverse normalization and inverse wavelet transformation on the final prediction result to obtain a prediction result related to the original power.
The specific neural network architecture in this embodiment is shown in table 1:
TABLE 1 neural network architecture
The above table gives the selection ranges of different neural network structure parameters, and the training parameter selection is as in table 2:
TABLE 2 training parameter selection
In the network training process, historical photovoltaic power data are normalized, and the normalization formula is
In the above-mentioned formula, the compound has the following structure,for the selected historical photovoltaic sequence to be used,is the minimum value in the sequence and is,the maximum value is obtained, and in order to evaluate the performance of the prediction model, the prediction result is evaluated by using MAE and RMSE in the embodiment
In the above formulaIn order to output the length of the sequence,is a firstThe number of the original values is set to,is a firstAnd (4) predicting the value.
Fig. 3 shows a specific case of a photovoltaic power smoothing strategy considering power prediction. Suppose to carvetThe upper and lower limits of the smooth value at this time are. If the residual capacity of the energy storage battery is enough, the calculated smooth value is within the limit. However, the smooth output result at this time will affect the smooth situation at the future time. Therefore, in this embodiment, the smoothed value of the power at this time is corrected on the basis of the power predicted value. The specific process is as follows: firstly, DWT-CNN-LSTM prediction model is adopted for predictiont+And (3) smoothing the predicted value in advance by adopting a TD algorithm according to the value at the moment 1, and calculating the energy storage change condition of the predicted value. Two extreme conditions exist in the state of charge of the energy storage batteryThe strategy discusses two different cases separately. When the state of charge at the future moment is smoothed to 90% according to the algorithm, it is indicated that the overcharge phenomenon may be caused at the future moment, so the energy storage battery is subjected to the discharge operation at the moment; similarly, if the state of charge of the energy storage battery reaches 10% at a future time, it indicates that the energy storage battery may be over-discharged at the future time, so that the energy storage battery needs to be charged at the present time. In the smoothing process, because of the limit of the power fluctuation condition, the smoothing value modified by the proposed strategy at this moment needs to meet the following requirements: (1) The corrected smooth value is satisfied as much as possibleWithin the range. (2) The corrected smooth value gives consideration to the fluctuation rate, and the output fluctuation rate is the same as that of the TD algorithm as far as possible.
After two requirements are considered, a power smoothing correction table is provided as a reference, and the reference values of smoothing correction are given for different situations by taking two limit working conditions of SOC >90% and SOC <10% as examples respectively, as shown in Table 3.
TABLE 3 smooth correction value Table
In Table 3JTo representtPower value of each point in time window,Representing the rated capacity of the photovoltaic array. Table 3 will discuss the correction values for different cases. With SOC (t + 1)>For example, if 90% of the cases areIn order not to excessively affect the smoothness, the intermediate value is selected during smoothing in consideration ofAs a reference value. When in useThen, can selectAs a power reference value, but ifAccording to the set power reference valueThe photovoltaic power is lower than the original photovoltaic output power at the moment, the charging and discharging state may change according to the original set value, and the release of the power at the moment is not facilitated, so the set value is changed into(ii) a When in useWhen the zero reference power isHowever, there is another caseThen the selected power reference value is modified to. According to the invention, on the basis of the TD algorithm smoothing result, a power prediction strategy is adopted for correction, so that the energy storage SOC is prevented from being in a limit state for a long time.
In this embodiment, the proposed smoothing strategy not only ensures the smoothness of the power smoothing value, but also can reasonably arrange the whole charging and discharging process as much as possible. For more precise smoothing, the present embodiment will use a power fluctuation evaluation index of a short time scale, which is as follows
In the formulaFor the photovoltaic power after the energy storage smoothing,is composed ofiAnd the output power of the photovoltaic array sampled at the moment. Compared with the traditional evaluation index, the evaluation index is a power fluctuation evaluation index of 10s level, and can reflect hidden power fluctuation. According to the grid-connection regulation, the power fluctuation of the grid connection is required to be controlled within the regulation, and the power fluctuation in the whole smoothing process is limited according to the above formula.
The smoothing algorithm used in this embodiment is a Tracking Differentiator (TD), TD algorithm.
The discrete form of a conventional tracking differentiator is shown as follows,
in a conventional tracking differentiatorFor tracking the original signal of the signal to be detected,for the derivative of the original signal, the system step size ishWherein the input signal isu,Andas in a system in discrete casesStep andand (5) carrying out the steps.
And the fastest tracking differentiator is a special type of tracking differentiator in the form of input to the system (1)uSet to a special non-linear formfhan,The expression is as follows:
wherein the content of the first and second substances,is composed offhanThe difference between the output and the input signal,is composed offhanAn estimate of the derivative of the input signal is made,fhanin which there are two parameter tracking factors to be adjustedA filter factorThe input signal passes throughfhanThereafter, a noise-free input tracking signal and an estimate of the derivative of the input signal may be obtained.fhanThe whole phase plane is divided into a linear region, a reachable region and the like, when the conditions are metIndicating that the current state is within the reachable region; in the same way if the condition is satisfiedIndicates that it is currently located in the linear regionAnd (4) the following steps. When the system state is outside the linear region, then this meansfhanWill be expressed asIn a linear region, while the states are located inside the linear region,fhanwill be asIn the form of (1). The boundary between the inner and outer linear regions is a boundary curve, and the adoption of a partitioned structure means that the input signal can be quickly tracked without overshoot. Of steepest-tracking differentiatorsfhanThe middle tracking factor is mainly used for adjusting the speed of the output tracking original signal of the steepest tracking differentiator, and the filtering factor is mainly used for adjusting the smoothness degree of the output.
In the specific implementation process of the embodiment, a CNN-LSTM neural network is adopted to predict photovoltaic power, a loss function adopts Mean Square Error (MSE), adam is adopted as an optimizer, a prediction model is realized by using Python3.7.0 and Tensorflow2.0.0, input data is divided into three sets, namely a training set, a verification set and a testing set, the proportion of each set is 1, the learning rate is selected to be 0.001, the training times are 100 times, and the batch size is 64. The adopted data is the operation data of a certain 7kW photovoltaic array in east China, the sampling time of photovoltaic data is from 9 am to 3 pm, in order to improve the prediction accuracy, power data of similar meteorological days in nearly fifteen days are selected according to the meteorological forecast condition of a prediction day, 17280 pieces of training set data, 2160 pieces of verification sets and 2160 pieces of test sets are selected.
Fig. 4-8 are graphs showing the comparative effect of the proposed strategy. As can be seen from fig. 5, power smoothing is not possible due to the energy storage SOC reaching a limit around 12. Two places I and II are mainly explained, the SOC of the place I reaches 90 percent due to energy storage, and the battery still needs to be charged for achieving smoothness after the place I. Electric energy is released in advance at the position I after a power prediction strategy is adopted, and a charging space is reserved for the later moment; and in the second place, because more power is released in the previous stage, the battery has the capability of smoothing, so the smoothing effect is better than that of a TD smoothing algorithm. In addition to the analysis of the smoothed power, the present embodiment focuses more on the SOC variation, and the SOC variation in the whole smoothing process is shown in fig. 6. As can be seen from fig. 7, the SOC of the energy storage battery reaches 10% when there are three power smoothing, and the power smoothing time periods are respectively 9. The details of the two points I and II are mainly 9. The SOC at i reaches around the minimum value in order to absorb as much power as possible and set the smooth reference value small. The power smoothing effect after the moment I is better than that of the original smoothing algorithm as can be seen from the smoothing curve; and II, absorbing part of electric energy to meet the requirement of subsequent discharge, wherein the change of the SOC in the smoothing process is shown in the figure 8.
In summary, the power smoothing method provided in this embodiment can reduce the time that the energy storage battery is in the limit SOC, thereby ensuring the service life of the energy storage battery.
Claims (8)
1. A photovoltaic power smoothing method considering power prediction is characterized by comprising the following steps:
s1: acquiring photovoltaic power historical data, and training a prediction model;
s2: acquiring photovoltaic power data of the inverter, and obtaining a power smooth value at the moment and an energy storage SOC value at the moment after smoothing by adopting a smoothing algorithm on the acquired data at the moment;
s3: predicting the photovoltaic power at the next moment based on the trained prediction model, and obtaining a power smooth value at the next moment and a smoothed energy storage SOC value at the next moment by adopting a smoothing algorithm;
s4: judging whether the energy storage SOC value at the next moment is in a limited range,
if yes, the power smooth value and the energy storage SOC value at the moment are output,
if not, the power smooth value at the moment is corrected.
2. The photovoltaic power smoothing method considering power prediction as claimed in claim 1, wherein the prediction model uses multi-order wavelet transform to divide the photovoltaic power history data into a plurality of subsequences including approximate components and detail components, different CNN-LSTM neural networks are used for prediction for different subsequences, and the final prediction result is subjected to inverse normalization and inverse wavelet transform to obtain the prediction result about power.
3. The photovoltaic power smoothing method considering power prediction as claimed in claim 1, wherein in step S4, when the energy storage SOC value reaches the upper limit after being smoothed according to the algorithm at the next moment, the energy storage battery is discharged and operated at the current moment, and when the energy storage SOC value reaches the lower limit after being smoothed according to the algorithm at the next moment, the energy storage battery is charged and operated at the current moment.
4. The photovoltaic power smoothing method considering power prediction as claimed in claim 3, wherein in step S4, the energy storage SOC value at the next moment defines an upper line as 90% and a lower line as 10%.
5. The photovoltaic power smoothing method considering power prediction as claimed in claim 3, wherein in step S4, when the power smoothing value at the present moment is corrected, the smoothing correction reference value is as follows:
is provided withJRepresenttPower value of each point in time window,Representing the rated capacity of the photovoltaic array, SOC (t + 1) representing the energy storage SOC value at the next moment, and the upper limit and the lower limit of the defined SOC (t + 1) are respectively 90 percent and 10 percent,
SOC (t + 1) > 90%:
When Min: (J) +10% P u <Max(J)-10%P u And Max: (J)-10%P u >P g (t) When the smoothed modified reference value is Max: (J)-10%P u ;
When Min: (J) +10% P u <Max(J)-10%P u And Max: (J)-10%P u <P g (t) When the reference value is smoothly corrected to be;
When SOC (t + 1) < 10%:
when the temperature is higher than the set temperature Max (Max)J)-10%P u >Min(J)+10%P u The smooth correction reference value is Min (J)+10%P u ;
When Min: (J) <Max(J)-10%P u <Min(J) +10%P u And Max: (J)-10%P u <P g (t) When the smoothed modified reference value is Max: (J)-10%P u ;
When Min: (J) <Max(J)-10%P u <Min(J) +10%P u And Max: (J)-10%P u >P g (t) The smooth correction reference value is Min (J);
When Max (J)-10%P u < Min(J) And Min: (J)<P g (t) The smooth corrected reference value is Min: (J);
When Max (J)-10%P u < Min(J) And Min: (J)>P g (t) When the smoothed correction reference value is Max: (J) -10%P u 。
6. The photovoltaic power smoothing method considering power prediction as claimed in claim 1, wherein during power smoothing, a power fluctuation evaluation index of 10s level in a short time scale is adopted, and the evaluation index is as follows:
7. The photovoltaic power smoothing method considering power prediction as claimed in claim 6, wherein in step S2, the sampling time interval is set to 10S, and the communication protocol adopts a Modbus communication protocol.
8. The method of photovoltaic power smoothing with power prediction taken into account as claimed in any one of claims 1-7, wherein said smoothing algorithm employs a steepest-tracking differentiatorThe fastest tracking differentiator is the input of the conventional tracking differentiatoruSet to a special non-linear formfhan,The expression is as follows:
wherein, the first and the second end of the pipe are connected with each other,is composed offhanThe difference between the output and the input signal,is composed offhanAn estimate of the derivative of the input signal is made,fhanin which there are two parameter tracking factors to be adjustedFilter factorInput signal is passed throughfhanThereafter, a noise-free input tracking signal and an estimate of the derivative of the input signal may be obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211352600.5A CN115411749A (en) | 2022-11-01 | 2022-11-01 | Photovoltaic power smoothing method considering power prediction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211352600.5A CN115411749A (en) | 2022-11-01 | 2022-11-01 | Photovoltaic power smoothing method considering power prediction |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115411749A true CN115411749A (en) | 2022-11-29 |
Family
ID=84167871
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211352600.5A Pending CN115411749A (en) | 2022-11-01 | 2022-11-01 | Photovoltaic power smoothing method considering power prediction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115411749A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008295208A (en) * | 2007-05-24 | 2008-12-04 | Kawasaki Heavy Ind Ltd | Method and apparatus for smoothing power and method for designing this apparatus |
CN103986190A (en) * | 2014-05-26 | 2014-08-13 | 电子科技大学 | Wind and solar storage combining power generation system smooth control method based on power generation power curves |
CN105552969A (en) * | 2015-12-29 | 2016-05-04 | 北京国电通网络技术有限公司 | Power prediction-based distributed output power smoothing method and system for photovoltaic generation |
CN109787260A (en) * | 2019-01-31 | 2019-05-21 | 中国电力科学研究院有限公司 | A kind of energy storage system control method and system of the photovoltaic power generation short-term fluctuation of smooth distribution formula |
WO2020215405A1 (en) * | 2019-04-25 | 2020-10-29 | 山东大学 | Photovoltaic power station power prediction method and system based on grid-tie inverter operating data |
-
2022
- 2022-11-01 CN CN202211352600.5A patent/CN115411749A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2008295208A (en) * | 2007-05-24 | 2008-12-04 | Kawasaki Heavy Ind Ltd | Method and apparatus for smoothing power and method for designing this apparatus |
CN103986190A (en) * | 2014-05-26 | 2014-08-13 | 电子科技大学 | Wind and solar storage combining power generation system smooth control method based on power generation power curves |
CN105552969A (en) * | 2015-12-29 | 2016-05-04 | 北京国电通网络技术有限公司 | Power prediction-based distributed output power smoothing method and system for photovoltaic generation |
CN109787260A (en) * | 2019-01-31 | 2019-05-21 | 中国电力科学研究院有限公司 | A kind of energy storage system control method and system of the photovoltaic power generation short-term fluctuation of smooth distribution formula |
WO2020215405A1 (en) * | 2019-04-25 | 2020-10-29 | 山东大学 | Photovoltaic power station power prediction method and system based on grid-tie inverter operating data |
Non-Patent Citations (2)
Title |
---|
赵海成: "考虑功率预测的光伏功率平滑策略研究", 《电力电子技术》 * |
赵海成等: "SA-PSO优化边界最速跟踪微分器在光伏功率平滑中的应用", 《电子测量技术》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112217202B (en) | Distributed new energy, energy storage and power distribution network planning method considering flexibility investment | |
CN107957562B (en) | Online prediction method for residual life of lithium ion battery | |
CN105846461B (en) | Control method and system for large-scale energy storage power station self-adaptive dynamic planning | |
CN114372417A (en) | Electric vehicle battery health state and remaining life evaluation method based on charging network | |
CN110658459B (en) | Lithium ion battery state of charge estimation method based on bidirectional cyclic neural network | |
CN107766986A (en) | Leak integral form echo state network on-line study photovoltaic power Forecasting Methodology | |
CN104951846A (en) | Microgrid short-term power and load prediction system and dynamic error classification and correction method | |
CN112103980A (en) | Energy management method of hybrid energy storage system combining AGC frequency modulation of thermal power generating unit | |
CN112858920B (en) | SOC estimation method of all-vanadium redox flow battery fusion model based on adaptive unscented Kalman filtering | |
CN105207243A (en) | Battery energy capacity management method for real-time power prediction and correction of wind power plant | |
CN112736973A (en) | Battery energy storage capacity configuration method and system for stabilizing output fluctuation of wind power and photovoltaic power station | |
CN116500475B (en) | Energy storage acquisition method and system with real-time SOC correction compensation | |
CN115411749A (en) | Photovoltaic power smoothing method considering power prediction | |
CN106849138B (en) | A kind of energy storage configuration method based on butterworth filter | |
CN116885691A (en) | Wind power climbing event indirect prediction method | |
CN108631368B (en) | Energy storage configuration method considering wind storage system joint scheduling under energy storage operation loss | |
CN113779861B (en) | Photovoltaic Power Prediction Method and Terminal Equipment | |
CN115663853A (en) | Hybrid energy storage unbalanced power distribution method based on real-time empirical mode decomposition | |
CN115036931A (en) | Active power grid reactive voltage affine adjustable robust optimization method and device | |
CN115764855A (en) | Real-time adjustable capacity and available electric quantity prediction method for electric vehicle quick charging station | |
CN111160653B (en) | Distributed energy storage system wind power consumption capability monitoring method based on cloud computing | |
CN112114255A (en) | Lithium battery state of charge estimation method based on combined ensemble Kalman filtering | |
CN110751342B (en) | Power load time series prediction method based on MBG optimization | |
CN112216883B (en) | Lithium battery equalization method, system, equipment and storage medium | |
CN112798974B (en) | Storage battery SOH on-line monitoring method and system for isolated base station hybrid power supply system |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20221129 |
|
WD01 | Invention patent application deemed withdrawn after publication |