CN117394439A - Method for improving power generation self-utilization rate of energy storage system - Google Patents
Method for improving power generation self-utilization rate of energy storage system Download PDFInfo
- Publication number
- CN117394439A CN117394439A CN202311635515.4A CN202311635515A CN117394439A CN 117394439 A CN117394439 A CN 117394439A CN 202311635515 A CN202311635515 A CN 202311635515A CN 117394439 A CN117394439 A CN 117394439A
- Authority
- CN
- China
- Prior art keywords
- data
- energy storage
- power generation
- model
- power
- 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
- 238000004146 energy storage Methods 0.000 title claims abstract description 67
- 238000010248 power generation Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 76
- 238000012549 training Methods 0.000 claims abstract description 59
- 230000005611 electricity Effects 0.000 claims abstract description 55
- 238000011217 control strategy Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 34
- 238000012545 processing Methods 0.000 claims description 23
- 238000010606 normalization Methods 0.000 claims description 20
- 238000005070 sampling Methods 0.000 claims description 19
- 238000006243 chemical reaction Methods 0.000 claims description 15
- 238000009499 grossing Methods 0.000 claims description 15
- 230000009471 action Effects 0.000 claims description 14
- 238000005457 optimization Methods 0.000 claims description 11
- 230000002787 reinforcement Effects 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000004140 cleaning Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 claims description 7
- 238000013112 stability test Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 4
- 238000007619 statistical method Methods 0.000 claims description 4
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 claims 4
- 230000008569 process Effects 0.000 description 5
- 238000013499 data model Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000012217 deletion Methods 0.000 description 3
- 230000037430 deletion Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
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
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/381—Dispersed generators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/092—Reinforcement learning
-
- 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
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- 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
-
- 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
- H02J2300/26—The renewable source being solar energy of photovoltaic origin involving maximum power point tracking control for photovoltaic sources
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention is suitable for the technical field of new photovoltaic energy, and provides a method for improving the power generation self-utilization rate of an energy storage system, which comprises the following steps: s1, collecting historical electricity utilization data of a user by a system and preprocessing the electricity utilization data; s2, predicting the power generation power of the solar panel, optimizing an energy storage control strategy and predicting the load demand, and correspondingly selecting a learning algorithm; s3, training the historical electricity consumption data through a selected learning algorithm and establishing a prediction model of a power supply strategy; s4, verifying and evaluating the trained power supply strategy model by taking part of historical data as a test set; s5, automatically adjusting a power supply strategy according to the real-time power demand and the photovoltaic power generation condition according to the trained power supply strategy model system; and S6, feeding back and adjusting a power supply strategy in real time by the system according to the real-time power utilization condition and the photovoltaic power generation condition of the user. The method enables the electric quantity output by the solar panel to just cover the power of the load end, and the redundant electric quantity generated by the electric quantity is converged into the energy storage equipment through MPPT.
Description
Technical Field
The invention belongs to the technical improvement field of new photovoltaic energy, and particularly relates to a method for improving the power generation self-utilization rate of an energy storage system.
Background
Photovoltaic power generation technology and industry are not only an important supplement to the energy sources of the prior art, but also have the potential to become the main energy source in the future. Along with the development of science and technology and society, more and more energy sources are needed, the conventional energy sources are fewer, the utilization of renewable energy sources can save a lot of conventional energy sources, solar energy is taken as one of new energy sources, obvious advantages are occupied, the trend of development is long, solar energy is photovoltaic power generation or focused heating by utilizing light energy and the like, the advantages of energy conservation, environmental protection and the like can be achieved, at present, the development of solar photovoltaic power generation is rapid, and various photovoltaic energy storage systems frequently and continuously promote industry to advance.
The existing photovoltaic energy storage system comprises the following components: photovoltaic module, dc-to-ac converter module, energy storage module and electric wire netting. The photovoltaic module is used for converting solar energy into electric energy, and the inverter module is used for converting single direct current of the photovoltaic module or the energy storage module into three alternating currents; and boosting the voltage of the input end; the power grid is connected when the photovoltaic module and the energy storage module do not output electric energy.
Disclosure of Invention
The invention aims to provide a method for improving the power generation self-utilization rate of an energy storage system, which aims to solve the technical problems that in the existing photovoltaic energy storage system, electricity generated by a photovoltaic PV panel is directly supplied to a household power grid through a micro inverter, but the micro inverter is usually provided with fixed output power, and when household electricity does not occupy the output power of the micro inverter, redundant electric quantity automatically enters the power grid to be wasted.
The invention is realized in such a way, a method for improving the power generation self-utilization rate of the energy storage system comprises the following steps:
s1, collecting historical electricity utilization data of a user by a system and preprocessing the electricity utilization data;
s2, predicting the power generation power of the solar panel, optimizing an energy storage control strategy and predicting the load demand, and correspondingly selecting a learning algorithm;
s3, training the historical electricity consumption data through a selected learning algorithm and establishing a prediction model of a power supply strategy;
s4, verifying and evaluating the trained power supply strategy model by taking part of historical data as a test set;
s5, automatically adjusting a power supply strategy according to the real-time power demand and the photovoltaic power generation condition according to the trained power supply strategy model system;
and S6, feeding back and adjusting a power supply strategy in real time by the system according to the real-time power utilization condition and the photovoltaic power generation condition of the user.
The invention further adopts the technical scheme that: the step S1 further comprises the following steps:
s11, cleaning historical electricity consumption data collected by the system;
s12, converting and normalizing the historical electricity consumption data;
s13, carrying out data smoothing and sampling processing on the historical electricity consumption data;
s14, extracting and classifying the characteristics of the historical power utilization data subjected to conversion and normalization processing or data smoothing and sampling processing;
s15, dividing the historical electricity consumption data processed by the method into a training set and a testing set.
The invention further adopts the technical scheme that: the step S11 includes the following steps:
s111, deleting repeated data points;
s112, filling or deleting the data points with the missing values through an interpolation method;
s113, identifying and processing the abnormal data by using a statistical method or an abnormal detection algorithm.
The invention further adopts the technical scheme that: the step S12 includes the following steps:
s121, reducing scale difference of exponential growth or attenuation trend data through logarithmic conversion;
s122, carrying out maximum-minimum normalization and standardization treatment on the data.
The invention further adopts the technical scheme that: in step S13, the average value of the electricity consumption data over a period of time is calculated to reduce the fluctuation and the sampling frequency.
The invention further adopts the technical scheme that: in the step S15, 80% of the data set is divided into training sets, and 20% is divided into test sets.
The invention further adopts the technical scheme that: in the step S2, a time sequence prediction algorithm is adopted for the prediction selection regression algorithm, the energy storage control strategy optimization selection reinforcement learning algorithm and the load demand prediction of the solar panel.
The invention further adopts the technical scheme that: the step S3 further comprises the following steps:
s31, initializing model parameters according to a learning algorithm selected by each module;
s32, importing the data of the divided training set and the data of the test set into the model, and training the model by using the training set data.
The invention further adopts the technical scheme that: the prediction selection regression algorithm of the solar panel generated power comprises the following steps:
s211, randomly initializing parameters;
s212, forward propagation;
s213, calculating loss;
s214, back propagation;
s215, repeatedly executing the steps S212-S214 to adjust the model to an optimal state to complete training;
the energy storage control strategy optimally selects a reinforcement learning algorithm, a Q-learning algorithm is used, and the Q-learning algorithm updates a Q value function through iteration until the Q value function converges to an optimal solution, and the method comprises the following steps of:
s221, initializing a Q value function;
s222, interacting with the environment;
s223, updating the Q value function;
s224, repeatedly executing the steps S222-S223 until the Q value function converges or reaches the designated training times;
s225, an optimal energy storage control strategy is formulated by selecting actions with the maximum Q values in each state.
The invention further adopts the technical scheme that: the load demand prediction adopts a time sequence prediction algorithm to carry out load prediction by using an ARIMA algorithm, and comprises the following steps:
s231, carrying out stability test on the load demand data;
s232, selecting parameters of an ARIMA model according to an auto-correlation diagram (ACF) and a partial auto-correlation diagram (PACF);
s233, training the historical load demand data according to the selected ARIMA model parameters to obtain an ARIMA model.
The beneficial effects of the invention are as follows: in the method, the power output to the load end by the micro inverter is automatically regulated in real time, so that the power output to the load end by the micro inverter from the electric quantity emitted by the solar panel just can cover the power of the load end, and the redundant electric quantity generated by the solar panel is converged into the energy storage equipment through MPPT.
Drawings
FIG. 1 is a flow chart of a method for improving the power generation self-utilization of an energy storage system according to an embodiment of the present invention.
Fig. 2 is a block diagram of a hardware system provided by an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative and intended to explain the present invention and should not be construed as limiting the invention.
In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
As shown in fig. 1, a flowchart of a method for improving the power generation self-utilization rate of an energy storage system provided by the invention is as follows:
and (3) adopting a preset algorithm to adjust the power output by the MPPT to the micro inverter, so that the power output by the micro inverter just covers the power of a load end, and therefore, the electric energy generated by the photovoltaic PV panel covers the total power output of the family and then the surplus electric energy can be gathered into the energy storage equipment to improve the power generation self-utilization rate of the whole energy storage system.
1. MPPT obtains user historical electricity consumption data, and establishes an electricity consumption model according to the user historical electricity consumption data.
2. The user sets the power output by the micro inverter to the load according to the self demand and sets the running mode of the energy storage system. Typically, the power of the micro-inverter may be autonomously set in steps of 100W within 800W, or set to an adaptive power mode. The operating mode of the energy storage system may generally be set to one of a low energy consumption mode, a standard mode, and a high performance mode.
As shown in fig. 2, the hardware system includes a solar panel, a controller, an energy storage device, a micro inverter, a power grid, a load AND an MPPT AND Pvhub (optimizer), wherein the output end of the solar panel outputs the converted electric energy to the input end of the MPPT AND Pvhub, the output end of the MPPT AND Pvhub outputs the energy to the micro inverter, the micro inverter AND the power grid respectively output the energy in two directions, the micro inverter also outputs the energy to the load, the MPPT AND Pvhub is connected with the controller, AND the MPPT AND Pvhub AND the energy storage device output the energy in two directions. The photovoltaic panel power generation system has the MPPT algorithm function of maximum point power tracking and the function of simulating a photovoltaic curve, and electric energy generated by the photovoltaic panel can be directly fed into the energy storage equipment through the optimizer.
The following steps are executed in the hardware system:
step S1, collecting historical electricity consumption data of a user by a system and preprocessing the electricity consumption data; the system firstly needs to collect historical electricity consumption data of users, including information such as electricity consumption, electricity consumption time, peak-valley time periods and the like. Such data may be obtained through a smart meter, an energy monitoring system, or a user-authorized data source.
The collected historical electricity utilization data is preprocessed and arranged so as to facilitate subsequent learning and analysis. This may include processing steps such as data cleansing, outlier removal, data normalization, etc.
Data cleaning: and cleaning the collected historical electricity consumption data to remove possible errors, deletions or abnormal values. The method comprises the following steps: duplicate data points are deleted. Reason 1 for repeated data points, data acquisition error: in the data acquisition process, data repetition may be caused due to sensor faults, communication errors, data transmission problems and the like. For example, a data point is repeatedly recorded or transmitted multiple times. 2. Data storage or processing errors: during data storage or processing, errors may occur, resulting in the data being repeatedly saved or processed. This may be due to a program error, database problem, or erroneous steps in the data processing flow. 3. Data merge errors: data duplication may occur when data collected from different data sources or different time periods are combined. This may be due to overlap between data sources, timestamp problems, or data merge algorithm errors. 4. Human error: human error may also cause data duplication. For example, when data is manually input, duplication of data may result from repeated typing or copy-and-paste errors. And detecting and processing the missing values, and filling or deleting the data points with the missing values by interpolation. Interpolation is performed by polynomial interpolation: fitting a curve by the acquired sampling points, and then interpolating at the missing points, wherein the value is positioned on the fitted curve; deletion of missing values: in some cases, if the number of missing values is small or the missing values cannot be reasonably interpolated, the data point where the missing value is located may be selected for deletion. This may result in a reduction in the amount of data, but may ensure the integrity and consistency of the data. Abnormal values are detected and processed, and abnormal data may be identified and processed using statistical methods or abnormality detection algorithms.
1. Data conversion and normalization: and converting and normalizing the historical electricity utilization data to eliminate dimensional differences among different features, and ensuring that the data are compared and analyzed on the same scale. The main purpose of data conversion and normalization is: eliminating dimension difference: the measurement units of different features may be different, and the range of values of the data may also be greatly different. For example, in a photovoltaic energy storage system, the unit of electricity consumption and time, the magnitude of the difference are large. Such dimensional differences can affect the data analysis and training results of the model. Through data conversion and normalization, the numerical values of different features can be adjusted to similar magnitudes, and the influence of dimensional differences among the features on a result is avoided.
2. Stability of data processing is improved: some machine learning algorithms are sensitive to the scale of the data, especially algorithms that involve distance metrics (e.g., K-means clusters, KNN classifiers, etc.). If the scale difference between features is large, it may result in some features having too much effect in the model, while the contribution of other features is ignored. Through normalization, the relative balance of the contribution of the features in the model can be ensured, and the stability and reliability of the model are improved.
3. The convergence speed of the model is accelerated: in some optimization algorithms, such as gradient descent, the scale of the data can affect the convergence rate of the algorithm. If the scale of some features is large, the gradient update may be small, resulting in slow algorithm convergence. Through normalization, the gradient range can be kept consistent, and the convergence rate of the optimization algorithm is accelerated.
The conversion and normalization method comprises the following steps: logarithmic conversion: for data that has an exponential growth or decay trend, logarithmic transformation may be employed to reduce the scale difference of the data. Max-Min normalization (Min-Max normalization): the data is mapped linearly to a specific range, typically between 0 and 1, preserving the relative order of the original data. Normalization (Z-score normalization): the data were converted to a distribution with a mean of 0 and standard deviation of 1 by subtracting the mean and dividing by the standard deviation.
Data smoothing and sampling: in some cases, historical electricity usage data may be subject to noise or high frequency fluctuations. The purpose of this step is: 1. noise filtering: data smoothing may help filter noise or outliers in the data to obtain cleaner and reliable data. This is important for improving the data quality and reliability of the analysis results. 2. Data compression: for large-scale data sets, data smoothing and sampling can reduce the amount of data, reducing the computational and storage requirements. This is particularly important in resource constrained environments. 3. Reducing complexity: some data sets may contain high frequency fluctuations or too detailed information, resulting in increased complexity of the data. By data smoothing and sampling, the data can be converted into a more simplified and easily understood form. To eliminate noise and reduce the complexity of the data, the method of data smoothing and sampling is adopted: moving average: an average over a period of time is calculated to reduce the volatility of the data. Downsampling: the sampling frequency of the data is reduced, for example from the minute level to the hour level, to reduce the amount of data and retain sufficient information.
The data conversion and normalization and the data smoothing and sampling are parallel steps, and are not sequential steps, after the data are cleaned, the cleaned data are subjected to data conversion and normalization to obtain a data model 1, and the cleaned data are subjected to data smoothing and sampling to obtain a data model 2.
Feature extraction (feature classification): useful features (classification categories) are extracted from the preprocessed data (data model 1 or data model 2) for subsequent modeling and learning. Useful features include: average electricity consumption: the average power usage per day, week or month is calculated. Peak Gu Chayi: the difference between the peak power usage and the valley power usage per day or week is calculated. Time correlation: features such as time trend, periodicity or correlation of electricity consumption are considered. Weather influence: if weather data exists, the influence of weather factors on the electricity consumption of the user, such as temperature, humidity and the like, can be extracted. Holiday effects: the influence of holidays on electricity consumption of users is extracted to distinguish electricity consumption modes of workdays and rest days. Photovoltaic power generation amount: if photovoltaic power generation data exist, the characteristic of the photovoltaic power generation can be extracted for comparison and analysis with the electricity consumption of a user. Trend of energy consumption: and analyzing the long-term trend of the electricity consumption of the user, and predicting the future electricity consumption requirement. Energy storage state: if an energy storage system (energy storage battery) is present, characteristics of the energy storage state may be extracted to monitor and adjust the power supply strategy of the energy storage system. User behavior characteristics: such as the user's power usage habits, power saving actions, etc., which may affect the user's power usage patterns.
Dividing data: the data after preprocessing and feature extraction are divided into a training set and a testing set. Typically, a portion of the data is used for training and parameter tuning of the model, while the remaining portion is used as a test set to evaluate the performance and generalization ability of the model. 80% of the dataset was used as training set and the remaining 20% was used as test set.
Step S2, a learning algorithm is correspondingly selected for prediction of the solar panel power generation power, optimization of an energy storage control strategy and prediction of load demand; and (3) learning algorithm selection: 1. and a regression algorithm is selected for the prediction of the solar panel power generation, and the regression algorithm has higher accuracy in predicting the target variable of the continuous value. 2. And optimizing and selecting a reinforcement learning algorithm for the energy storage control strategy, wherein the reinforcement learning algorithm can find an optimal strategy and make a control decision. The scheme optimizes the charging and discharging strategy of the energy storage system by adopting Deep Q Network (DQN) algorithm so as to maximize the energy storage efficiency and the power supply quality. 3. For load demand prediction, a time sequence prediction algorithm is adopted, and load distribution of loads of the household photovoltaic energy storage system on a time sequence is usually determined, so that the time sequence prediction algorithm is adopted to have higher accuracy.
Step S3, training the historical electricity consumption data through a selected learning algorithm and establishing a prediction model of a power supply strategy; training the historical electricity utilization data by using a selected machine learning algorithm, and establishing a prediction model of a power supply strategy. The training process correlates historical electricity usage data with actual power supply strategies to learn the rules and patterns of the power supply strategies.
Model initialization, namely initializing model parameters according to a selected learning algorithm. The initial parameters may be adjusted using default parameters or empirically.
The data of the training set and the test set are input into the model, and the model is trained using the data of the training set. During training, the model compares the loss function (objective function) to the actual label, adjusting the parameters to minimize the loss function.
The following is an example of model training:
solar panel generated power prediction algorithm model: assuming a simple regression problem, the generated power of the solar power generation system is predicted by the training set. The training set contains input characteristics and corresponding real generated power, and the data are as follows:
a simple linear regression model was chosen for training, assuming the equation for the model is: y=w1×characteristic 1+w2×characteristic 2+b
Where w1 and w2 are weight parameters of the model and b is a bias parameter.
The training process is as follows:
1) Initialization parameters): the values of weights w1, w2 and bias b are randomly initialized.
2) Forward propagation): the input features in the training set are brought into the model, and the predicted value y_pred is calculated.
For the first sample (6 hours, weather condition 1), the predicted value is: y_pred1=w1×6+w2×1+b;
for the second sample (7 hours, weather condition 2), the predicted value is: y_pred2=w1×7+w2×2+b;
and so on for other samples.
3) Calculating the loss: and comparing the predicted value y_pred with the real generated power y_true in the training set, and calculating the value of the loss function.
For the first sample, the loss is: loss1= (y_pred1-50) ≡2;
for the second sample, the loss is: loss2= (y_pred2-55) ≡2;
and so on for other samples.
4) Back propagation): parameters of the model are adjusted according to the loss function by an optimization algorithm (such as a gradient descent method) so as to reduce the value of the loss function.
Calculating the partial derivative of the loss function on the parameter to obtain the gradient: grad_w1, grad_w2, grad_b.
Updating parameters according to gradient and learning rate (learning rate): w (w) n = w1 - learning_rate * grad_w1，w m = w2 - learning_rate * grad_w2，b n = b - learning_rate * grad_b;w n 、w m Is the weight parameter of the updated model, b n Is the updated bias parameter.
5) Repeating the steps 2-4): the forward propagation and the backward propagation are iterated until the loss function reaches a satisfactory small value or a specified number of training times.
6) And (3) training: when training is finished, parameters (w 1, w2 and b) of the model are adjusted to an optimized state, and the model is trained and can be used for prediction of the generated power.
In the training process, the optimization algorithm aims to continuously adjust model parameters so that the model can accurately predict the generated power, and therefore errors between a predicted value and an actual value are minimized. This is the training process of the model by which the model can learn the relationship between the generated power and the input features from the historical data and then be used to predict the generated power of the unknown data.
Optimization of the energy storage control strategy may be addressed using a reinforcement learning algorithm. In reinforcement learning, an energy storage controller (Energy Storage Controller) learns to formulate optimal control strategies by interacting with the environment (photovoltaic power generation system, load demand, etc.) to maximize a cumulative Reward (report) signal. The following is a specific example of reinforcement learning algorithm: q-learning;
given a simplified energy storage control problem, the goal is to optimize the charge and discharge strategy of the energy storage system to maximize the energy utilization of the system. The system has an energy storage, and each time can choose to charge, discharge or do no operation. At the same time, the environment of the system is affected by solar power generation and load demands. The state of the system may be represented by the following features:
1) Current battery state (remaining energy storage amount)
2) Current solar power generation
3) Current load demand)
The energy storage controller selects an action to adjust the state of the energy storage system at each instant in time based on the current state. The goal is to guide the taking of optimal actions in each state by learning a Q-value function. The Q-value function represents the expected jackpot that would be achieved by selecting action a in state s.
The core of the Q-learning algorithm is to update the Q-value function by iteration until the optimal solution is converged. The method comprises the following specific steps:
1) Initializing a Q-value function): for all possible states and combinations of actions, the Q value is initialized to 0 or a random value.
2) A), interact with the environment: the agent selects an action and executes it according to the current state and the Q-value function. The environment returns the next state, rewards, and signals of whether to terminate based on the actions and current state selected by the agent.
3) Updating the Q value function): the Q-factor is updated using a Q-learning algorithm based on the current state, the selected action, the prize, and the next state. The specific update formula is: q(s) n , a n ) =q (s, a) +α (r+γ×max (Q (s ', a')) -Q (s, a)), where Q (s n ,a n ) For updated Q values, Q (s, a) represents the Q value of the selected action a in state s, α is learning rate, r is the obtained prize, γ is discount factor, s ' is the next state, and a ' is all possible actions in s '.
4) A) termination condition: repeating the steps 2 and 3 until the Q value function converges or reaches the designated training times.
5) Control strategy selection): after training is completed, an optimal energy storage control strategy is formulated by selecting the action with the largest Q value in each state.
The load prediction uses a time series prediction algorithm, and the following is a specific example of the time series prediction algorithm: ARIMA (autoregressive integrated moving average model).
ARIMA is a classical method of time series prediction that combines three parts, autoregressive (AR), differential (I), and Moving Average (MA). The prediction of the ARIMA model is based on the value of the past time and the error of the past time.
Assume a set of historical load demand data as follows:
the load prediction using ARIMA algorithm is as follows:
1) And (3) checking the stability: and carrying out stability test on the load demand data to ensure that the data is stable (namely, the mean value and the variance do not change with time), and if the data is not stable, carrying out differential processing.
2) -parameters selection): parameters of the ARIMA model are selected based on an autocorrelation diagram (ACF) and a partial autocorrelation diagram (PACF). The autocorrelation map reflects the correlation of the time series with its own lag version, and the partial autocorrelation map reflects the linear correlation of the time series with its lag version.
3) Model training): and training the historical load demand data according to the selected ARIMA model parameters to obtain an ARIMA model.
4) Model prediction): and predicting the load demand in a future period of time by using a trained ARIMA model.
Examples: assuming that it is desired to predict load demand for the next 3 days, the ARIMA (1, 1) model was chosen. And training an ARIMA model according to the historical load demand data, and predicting.
The predicted results are as follows:
step S4, verifying and evaluating the trained power supply strategy model by taking part of historical data as a test set; and verifying and evaluating the trained power supply strategy model, and evaluating the prediction accuracy and performance of the model by using a part of historical data as a test set. If the model does not perform well, the steps of algorithm adjustment, feature extraction and the like are adjusted, and training and verification are conducted again.
Step S5, automatically adjusting a power supply strategy according to the real-time power demand and the photovoltaic power generation condition according to the trained power supply strategy model system; according to the trained model, the system can automatically adjust the power supply strategy according to the real-time power demand and the photovoltaic power generation condition. For example, when predicting peak electricity consumption by a user, the system may store enough electrical energy from the photovoltaic power generation and storage system in advance to meet peak electricity demand.
And S6, feeding back and adjusting a power supply strategy in real time by the system according to the real-time power utilization condition and the photovoltaic power generation condition of the user. The system should monitor the electricity consumption condition and the photovoltaic power generation condition of the user in real time, and adjust and feed back in real time according to the actual condition. This can help the system to continuously optimize the power supply strategy and make flexible adjustments according to actual needs.
Another object of the present invention is to provide a system for increasing the power generation rate of an energy storage system, the system for increasing the power generation rate of an energy storage system comprising
The collection processing module is used for collecting historical electricity utilization data of a user by the system and preprocessing the electricity utilization data;
the algorithm selection module is used for correspondingly selecting a learning algorithm for the prediction of the solar panel power generation, the optimization of the energy storage control strategy and the prediction of the load demand;
the training and establishing model module is used for training the historical electricity consumption data through the selected learning algorithm and establishing a prediction model of the power supply strategy;
the verification and evaluation module is used for verifying and evaluating the trained power supply strategy model by taking part of historical data as a test set;
the automatic adjustment module is used for automatically adjusting the power supply strategy according to the trained power supply strategy model system, the real-time power consumption requirement and the photovoltaic power generation condition;
and the real-time feedback adjustment module is used for feeding back and adjusting the power supply strategy in real time according to the real-time power utilization condition and the photovoltaic power generation condition of the user.
The collecting and processing module also comprises
The cleaning unit is used for cleaning the historical electricity data collected by the system;
the conversion and normalization unit is used for converting and normalizing the historical electricity consumption data;
the smoothing and sampling unit is used for carrying out data smoothing and sampling processing on the historical electricity consumption data;
the classification unit is used for extracting characteristic classification from the historical power utilization data subjected to conversion and normalization processing or data smoothing and sampling processing;
the data dividing unit is used for dividing the historical electricity consumption data processed by the processing unit into a training set and a testing set.
The cleaning unit comprises
A deleting unit configured to delete duplicate data points;
the filling unit is used for filling or deleting the data points with the missing values through an interpolation method;
and the exception handling unit is used for identifying and handling the exception data by using a statistical method or an exception detection algorithm.
The conversion and normalization unit comprises
A scale difference unit for reducing scale differences of the exponentially growing or decaying trend data by logarithmic conversion;
and the processing unit is used for carrying out maximum-minimum normalization and standardization processing on the data.
The smoothing and sampling unit reduces fluctuation and sampling frequency reduction of the power consumption data by calculating an average value in a period of time.
The data dividing unit divides 80% of the data set into training sets and 20% into test sets.
And a time sequence prediction algorithm is adopted for the prediction selection regression algorithm, the energy storage control strategy optimization selection reinforcement learning algorithm and the load demand prediction of the solar panel power generation in the algorithm selection module.
The training and creating model module also comprises
The initialization unit is used for initializing model parameters according to the learning algorithm selected by each module;
and the model training unit is used for importing the data of the divided training set and the test set into the model and training the model by using the training set data.
The prediction selection regression algorithm of the solar panel generated power comprises the following steps of
Randomly initializing parameters;
forward propagation;
calculating loss;
counter-propagating;
repeating forward propagation, calculation loss and backward propagation to adjust the model to an optimal state to complete training;
the energy storage control strategy optimally selects a reinforcement learning algorithm, a Q-learning algorithm is used, and the Q-learning algorithm updates a Q value function through iteration until the Q value function converges to an optimal solution, and the method comprises the following steps: initializing a Q value function, interacting with the environment, and updating the Q value function; repeatedly executing interaction with the environment, and updating the Q value function until the Q value function converges or reaches the designated training times; an optimal energy storage control strategy is formulated by selecting the action with the largest Q value in each state.
The load demand prediction adopts an ARIMA algorithm in a time sequence prediction algorithm to carry out load prediction, and comprises the following steps of
The stability test unit is used for carrying out stability test on the load demand data;
a model selection unit for selecting parameters of the ARIMA model according to an Autocorrelation Chart (ACF) and a Partial Autocorrelation Chart (PACF);
and the obtaining model unit is used for training the historical load demand data according to the selected ARIMA model parameters to obtain an ARIMA model.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (10)
1. The method for improving the power generation self-utilization rate of the energy storage system is characterized by comprising the following steps of:
s1, collecting historical electricity utilization data of a user by a system and preprocessing the electricity utilization data;
s2, predicting the power generation power of the solar panel, optimizing an energy storage control strategy and predicting the load demand, and correspondingly selecting a learning algorithm;
s3, training the historical electricity consumption data through a selected learning algorithm and establishing a prediction model of a power supply strategy;
s4, verifying and evaluating the trained power supply strategy model by taking part of historical data as a test set;
s5, automatically adjusting a power supply strategy according to the real-time power demand and the photovoltaic power generation condition according to the trained power supply strategy model system;
and S6, feeding back and adjusting a power supply strategy in real time by the system according to the real-time power utilization condition and the photovoltaic power generation condition of the user.
2. The method for increasing the power generation self-utilization rate of an energy storage system according to claim 1, wherein said step S1 further comprises the steps of:
s11, cleaning historical electricity consumption data collected by the system;
s12, converting and normalizing the historical electricity consumption data;
s13, carrying out data smoothing and sampling processing on the historical electricity consumption data;
s14, extracting and classifying the characteristics of the historical power utilization data subjected to conversion and normalization processing or data smoothing and sampling processing;
s15, dividing the historical electricity consumption data processed by the method into a training set and a testing set.
3. The method for increasing the power generation self-utilization rate of an energy storage system according to claim 2, wherein said step S11 includes the steps of:
s111, deleting repeated data points;
s112, filling or deleting the data points with the missing values through an interpolation method;
s113, identifying and processing the abnormal data by using a statistical method or an abnormal detection algorithm.
4. A method for increasing the power generation self-utilization rate of an energy storage system according to claim 3, wherein said step S12 comprises the steps of:
s121, reducing scale difference of exponential growth or attenuation trend data through logarithmic conversion;
s122, carrying out maximum-minimum normalization and standardization treatment on the data.
5. The method according to claim 4, wherein the step S13 is performed to reduce the fluctuation and the sampling frequency of the electricity consumption data by calculating an average value over a period of time.
6. The method according to claim 2, wherein in the step S15, 80% of the data set is divided into training sets and 20% is divided into test sets.
7. The method for improving the power generation self-utilization rate of the energy storage system according to claim 1, wherein the prediction selection regression algorithm, the energy storage control strategy optimization selection reinforcement learning algorithm and the load demand prediction of the power generation power of the solar panel in the step S2 adopt a time sequence prediction algorithm.
8. The method for increasing the power generation self-utilization rate of an energy storage system according to claim 1, wherein said step S3 further comprises the steps of:
s31, initializing model parameters according to a learning algorithm selected by each module;
s32, importing the data of the divided training set and the data of the test set into the model, and training the model by using the training set data.
9. The method of increasing the power generation self-utilization of an energy storage system of claim 7, wherein the predictive selection regression algorithm of the solar panel power generation comprises the steps of:
s211, randomly initializing parameters;
s212, forward propagation;
s213, calculating loss;
s214, back propagation;
s215, repeatedly executing the steps S212-S214 to adjust the model to an optimal state to complete training;
the energy storage control strategy optimization selection reinforcement learning algorithm uses a Q-learning algorithm, and the Q-learning algorithm iteratively updates a Q value function until the Q value function converges to an optimal solution, and the method comprises the following steps of:
s221, initializing a Q value function;
s222, interacting with the environment;
s223, updating the Q value function;
s224, repeatedly executing the steps S222-S223 until the Q value function converges or reaches the designated training times;
s225, an optimal energy storage control strategy is formulated by selecting actions with the maximum Q values in each state.
10. The method of increasing the power generation self-utilization of an energy storage system of claim 7, wherein the load demand prediction employs an ARIMA algorithm in a time series prediction algorithm for load prediction, comprising the steps of:
s231, carrying out stability test on the load demand data;
s232, selecting parameters of an ARIMA model according to the autocorrelation diagrams and the partial autocorrelation diagrams;
s233, training the historical load demand data according to the selected ARIMA model parameters to obtain an ARIMA model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311635515.4A CN117394439A (en) | 2023-12-01 | 2023-12-01 | Method for improving power generation self-utilization rate of energy storage system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311635515.4A CN117394439A (en) | 2023-12-01 | 2023-12-01 | Method for improving power generation self-utilization rate of energy storage system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117394439A true CN117394439A (en) | 2024-01-12 |
Family
ID=89441214
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311635515.4A Pending CN117394439A (en) | 2023-12-01 | 2023-12-01 | Method for improving power generation self-utilization rate of energy storage system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117394439A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018129829A1 (en) * | 2017-01-12 | 2018-07-19 | 沃太能源南通有限公司 | New type micro-network system and system-based networking and scheduling method |
| US20220373984A1 (en) * | 2021-05-19 | 2022-11-24 | Shandong University | Hybrid photovoltaic power prediction method and system based on multi-source data fusion |
| CN116596340A (en) * | 2023-05-22 | 2023-08-15 | 集美大学 | Ship comprehensive energy system energy management strategy with working condition adaptability |
| CN117013527A (en) * | 2023-08-08 | 2023-11-07 | 河北汉尧碳科新能科技股份有限公司 | Distributed photovoltaic power generation power prediction method |
-
2023
- 2023-12-01 CN CN202311635515.4A patent/CN117394439A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018129829A1 (en) * | 2017-01-12 | 2018-07-19 | 沃太能源南通有限公司 | New type micro-network system and system-based networking and scheduling method |
| US20220373984A1 (en) * | 2021-05-19 | 2022-11-24 | Shandong University | Hybrid photovoltaic power prediction method and system based on multi-source data fusion |
| CN116596340A (en) * | 2023-05-22 | 2023-08-15 | 集美大学 | Ship comprehensive energy system energy management strategy with working condition adaptability |
| CN117013527A (en) * | 2023-08-08 | 2023-11-07 | 河北汉尧碳科新能科技股份有限公司 | Distributed photovoltaic power generation power prediction method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN111275367B (en) | Regional comprehensive energy system energy efficiency state evaluation method | |
| CN112614009A (en) | Power grid energy management method and system based on deep expected Q-learning | |
| CN111382789B (en) | Power load identification method and system based on machine learning | |
| CN110969306B (en) | Power distribution low-voltage transformer area load prediction method and device based on deep learning | |
| CN112801388B (en) | Power load prediction method and system based on nonlinear time series algorithm | |
| CN116451873B (en) | Wind power generation power prediction method and system based on multi-scale double space-time network area | |
| CN116258269A (en) | Ultra-short-term load dynamic prediction method based on load characteristic decomposition | |
| CN115034485A (en) | Wind power interval prediction method and device based on data space | |
| CN113988398A (en) | Wind turbine generator power prediction method and device, electronic equipment and storage medium | |
| CN109165770A (en) | Novel photovoltaic power forecasting method based on AFSA-Elman | |
| CN113792939A (en) | Electric energy meter reliability prediction method and device based on mixed Weibull distribution | |
| CN115758151A (en) | Combined diagnosis model establishing method and photovoltaic module fault diagnosis method | |
| CN115859099A (en) | Sample generation method and device, electronic equipment and storage medium | |
| CN117851908B (en) | Improved on-line low-voltage transformer area electric energy meter misalignment monitoring method and device | |
| Zjavka | Power quality daily predictions in smart off-grids using differential, deep and statistics machine learning models processing NWP-data | |
| CN117974395A (en) | Environment-friendly community management method and system based on carbon emission monitoring | |
| CN111950752A (en) | Photovoltaic power station generating capacity prediction method, device and system and storage medium thereof | |
| Zhang et al. | Hybrid model for renewable energy and load forecasting based on data mining and EWT | |
| CN117394439A (en) | Method for improving power generation self-utilization rate of energy storage system | |
| CN116167465A (en) | Solar irradiance prediction method based on multivariate time series ensemble learning | |
| CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
| CN113111588A (en) | NO of gas turbineXEmission concentration prediction method and device | |
| CN117977580B (en) | Power line control method based on open source hong Meng system | |
| CN114219216B (en) | Operation scheduling method, system, device and medium of power 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 |