CN114707713A - Low-voltage distribution network load prediction method considering distributed power supply influence - Google Patents
Low-voltage distribution network load prediction method considering distributed power supply influence Download PDFInfo
- Publication number
- CN114707713A CN114707713A CN202210291236.XA CN202210291236A CN114707713A CN 114707713 A CN114707713 A CN 114707713A CN 202210291236 A CN202210291236 A CN 202210291236A CN 114707713 A CN114707713 A CN 114707713A
- Authority
- CN
- China
- Prior art keywords
- model
- ann
- data
- solution
- prediction
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 15
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000011156 evaluation Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012216 screening Methods 0.000 claims abstract description 7
- 238000012546 transfer Methods 0.000 claims abstract description 5
- 238000001514 detection method Methods 0.000 claims abstract description 4
- 238000012795 verification Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 30
- 238000005457 optimization Methods 0.000 claims description 11
- 238000010248 power generation Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000011056 performance test Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 239000002699 waste material Substances 0.000 claims description 3
- 229940116352 animax Drugs 0.000 claims description 2
- 230000000694 effects Effects 0.000 claims description 2
- 230000005611 electricity Effects 0.000 abstract description 6
- 230000008901 benefit Effects 0.000 abstract description 4
- 230000001502 supplementing effect Effects 0.000 description 8
- 239000013589 supplement Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000007405 data analysis Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 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
- 230000009467 reduction Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Game Theory and Decision Science (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computing Systems (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a low-voltage distribution network load prediction method considering distributed power supply influence, which comprises the following working steps of: the method comprises the following steps: selecting input and output variables based on user load data of a low-voltage power grid containing distributed photovoltaic; step two: screening and preprocessing data, and identifying unusual observation variables and detection trends; step three: dividing a data set into a training data set, a verification data set and a test data set; step four: selecting ANN model parameters, and determining the number of hidden layers, the number of layers, a transfer function and an evaluation standard; step five: training an ANN model; step six: realizing an ANN model and outputting a result; the invention has the advantages that the provided ANN-GROM prediction model with the external variables can effectively reduce prediction errors, and the ANN-GROM has a good prediction result for given household electricity data.
Description
Technical Field
The invention relates to the technical field of power distribution network protection, in particular to a low-voltage power distribution network load prediction method considering distributed power supply influence.
Background
With the development of economic society and the improvement of living standard of people, more and more families are using renewable energy, and with the world moving towards the future of clean energy and the new mode of walking into power load, the trend is increasingly obvious; however, the method also brings new and significant challenges to operation and maintenance of the power distribution network, such as increasing of unstable net load power utilization law and difficulty in prediction of low-voltage user load; the problem is that the user load rule of the current low-voltage distribution network and the power generation rule of the distributed power supply are lack of deep knowledge;
at present, most scholars mainly concentrate on load prediction research of a medium-high voltage power grid, and for load prediction of a low-voltage power distribution network, few scholars conduct research on load prediction considering the influence of a small distributed power supply of a low-voltage user, and a traditional ANN prediction model uses optimization technologies such as gradient descent and Gauss-Newton method to solve a learning algorithm and obtain the best performance in the ANN; furthermore, these conventional optimization techniques can be used to find the local optimal parameters of the ANN, requiring that the objective function simultaneously satisfies the following criteria: smoothness, continuity, and differentiability. However, these conventional optimization methods cannot be effectively used to optimize the parameters of the ANN prediction model because the power load has a high degree of uncertainty;
in view of the above, there is a need for an improved prediction model method that can accommodate the current need for uncertain prediction of electrical loads.
Disclosure of Invention
Because the existing ANN algorithm for load prediction is not strong in adaptability to power load and cannot meet the use requirements of people, a low-voltage distribution network load prediction method influenced by a distributed power supply is designed on the basis of the defects of the prior art, and the uncertainty of the low-voltage power load requirement can be well adapted through power generation of renewable energy sources.
The technical scheme of the invention is that the method for predicting the load of the low-voltage distribution network considering the influence of the distributed power supply comprises the following working steps:
the method comprises the following steps: selecting input and output variables based on user load data of a low-voltage power grid containing distributed photovoltaic;
step two: screening and preprocessing data, and identifying unusual observation variables and detection trends;
step three: dividing a data set into a training data set, a verification data set and a test data set;
step four: selecting ANN model parameters, and determining the number of hidden layers, the number of layers, a transfer function and an evaluation standard;
step five: training an ANN model;
step six: realizing an ANN model and outputting a result;
wherein, this technical scheme applies GROM in ANN model, and its implementation mode is based on Fibonacci function, namely
The series F0, F1, F2
Fn＝Fn-1+Fn-2 (1)
Where φ is called the golden ratio, for updating the search process and finding the optimal solution in two different phases, it includes the following steps:
the first step is as follows: calculating the average value of all possible solutions for training the ANN network; then considering the fitness function, comparing the mean solution with the worst solution, and replacing the worst solution with the mean solution if the mean solution results in a better fitness function value, the purpose of this process is to speed up the algorithm and get it converged as soon as possible.
The second step is that: in order to determine the search direction, a random solution and a mean solution are selected to be compared to explore the influence of the random solution and the mean solution on the search motion, and the method can help to determine and optimize ANN model parameters and can avoid deviation of a prediction model caused by selection of additional parameters.
To further supplement the technical scheme, the output variable in the step one is as follows: predicting load demand at time t + i
Photovoltaic output power
Input variables are: due to the strong robustness between weather conditions (e.g. temperature, wind speed) and output variables, external variables (e.g. temperature, wind speed) will be taken as important input variables, and in addition due to the important impact of the historical load demand of the home and the historical output power of the photovoltaic generation on the prediction results will also be taken as input variables.
Further supplementing the technical scheme, the step two of data screening and preprocessing comprises checking all data to avoid data waste.
Further supplementing the technical scheme, dividing the data set in step three: the processed data set is divided into training, validation and test data sets.
The technical scheme is further supplemented by the following steps of selecting parameters of an ANN model: the calculations to find and mitigate complex correlations are the reason for using the parametric functions.
In addition to the technical scheme, the number of the hidden layers and the number of the neurons are determined by a trial and error method.
To further supplement the technical solution, in order to improve the performance of the prediction model and reduce the peak value of high error to the maximum extent, external variables such as weather conditions are considered, and the influence of the external variables such as weather conditions on the ARIMAX and ANN prediction models is considered here, and the prediction model is first divided into sub-models to evaluate the influence of the external variables in the ani and ARIMAX models, specifically as follows:
model a 1: ARIMAX has two external variables (X)1Temperature, X2-wind speed);
model a 2: ARIMAX has an external variable (X)1-temperature);
model a 3: ARIMAX has an external variable (X)2-wind speed);
model a 4: ARIMAX only has photovoltaic power generation and user load and does not have a model of an external variable;
model NN 1: the ANN model has several external variables (X)1Temperature, X2Wind speed, X3Hour of the day, X4Previous hour data, X5-data of the same hour of the previous day);
model NN 2: ANN models do not contain weather Condition variables (X)3Hour of the day, X4Previous hour data, X5-data of the same hour of the previous day);
model NN 3: ANN model has weather variables (X) only1Temperature, X2-wind speed).
The technical scheme is further supplemented, parameters of the ANN prediction model are optimized by adopting a GROM algorithm, and the method can be realized by the following steps:
the method comprises the following steps: taking some random learning parameters of the ANN prediction model as population initialization parameters, and calculating the average value of the population;
step two: evaluating the fitness of each model parameter by adopting a learning cost function in the ANN, then comparing the fitness of the population mean solution with the fitness of the worst solution, and replacing the worst solution with the population mean solution if the mean solution obtains a better fitness function value, wherein the process in the GROM aims to enhance the optimization speed to make the algorithm converge;
step three: creating a random solution vector in the group to determine and specify the direction of the next advance and the size of the motion; the optimal parameter solution is the solution with the minimum objective function value, and in the GROM, the parameter solution needs to be updated and moves towards the direction of the population optimal solution.
Further supplementing the present solution, the accuracy of the prediction model is determined by using the Mean Absolute Percentage Error (MAPE) and the Root Mean Square Error (RMSE), as shown in equations (4) and (5);
where l (t) is actual data, such as actual user load data;
is predicted data; t is the current time step and T is the total number of time steps (observations).
Further supplementing the technical scheme, the performance test of the prediction model adopts the following steps:
the method comprises the following steps: comparing the prediction model performance of different data profiles:
step two: the influence of the external variables on the predictive model is evaluated.
The method has the advantages that the ANN-GROM prediction model is adopted to carry out different low-voltage prediction applications, the performance of the prediction model is improved by adopting external variables such as weather conditions and the like and an optimization method of GROM, the proposed prediction model is trained and tested by using real-time power grid data, and the prediction result of the prediction model shows that the proposed ANN-GROM prediction model with the external variables can effectively reduce prediction errors, particularly, for given household electricity data, the ANN-GROM has good prediction results and is superior to the traditional ANN and ARIMAX; with the benefit of reducing prediction errors, it is also possible to potentially learn the demand of low voltage grid users for electricity and to derive considerable technical and economic benefits from domestic electricity demand and photovoltaic power generation data analysis and the prediction model.
Drawings
FIG. 1 is an actual and predicted value of a single household load demand of the present invention;
FIG. 2 is an overview of actual and predicted payload demands of a single household user of the present invention;
Detailed Description
Because the existing ANN algorithm for load prediction is not strong in adaptability to power load and cannot meet the use requirements of people, a low-voltage distribution network load prediction method influenced by a distributed power supply is designed on the basis of the defects of the prior art, and the uncertainty of the low-voltage power load requirement can be well adapted through power generation of renewable energy sources. The patent provides an artificial neural network prediction model based on a golden ratio optimization method, and influences of external factors such as weather conditions on accuracy of the prediction model are considered. Firstly, analyzing the influence of different external variables on the accuracy of a prediction model, then establishing a prediction model aiming at the short-term net load of a household user and the output power of a photovoltaic system based on the external variables with obvious influence on load prediction and photovoltaic power generation prediction, and determining and optimizing ANN model parameters by adopting a Golden Ratio Optimization (GROM) method to reduce the prediction error of the model. The prediction result shows that compared with the traditional ANN model, the ANN-GROM model has better prediction effect.
In this patent, hourly temperature and historical data correlations are used in a photovoltaic prediction model. The prediction model not only provides net load prediction data of the power grid in the future day, but also predicts the user load and the distributed photovoltaic respectively and provides a corresponding prediction result.
In order to make the technical solution more clear to those skilled in the art, the technical solution of the present invention will be explained in detail below:
a low-voltage distribution network load prediction method considering distributed power supply influence comprises the following working steps:
the method comprises the following steps: selecting input and output variables based on user load data of a low-voltage power grid containing distributed photovoltaic;
step two: screening and preprocessing data, and identifying unusual observation variables and detection trends;
step three: dividing a data set into a training data set, a verification data set and a test data set;
step four: selecting ANN model parameters, and determining the number of hidden layers, the number of layers, a transfer function and an evaluation standard;
step five: training an ANN model;
step six: realizing an ANN model and outputting a result;
in the technical scheme, GROM is applied to an ANN model, and the implementation mode is based on a Fibonacci function, namely
The series F0, F1, F2
Fn＝Fn-1+Fn-2 (1)
Where phi is called the golden ratio, for updating the search process and finding the optimal solution in two different stages, which comprises the following steps:
the first step is as follows: calculating the average value of all possible solutions for training the ANN network; then considering the fitness function, comparing the mean solution with the worst solution, and replacing the worst solution with the mean solution if the mean solution results in a better fitness function value, the purpose of this process is to speed up the algorithm and get it converged as soon as possible.
The second step is that: in order to determine the search direction, a random solution and a mean solution are selected to be compared to explore the influence of the random solution and the mean solution on the search motion, and the method can help to determine and optimize ANN model parameters and can avoid deviation of a prediction model caused by selection of additional parameters.
To further supplement the technical scheme, the variable is output in the step one: predicting load demand at time t + i
Photovoltaic output power
Input variables are: due to the strong robustness between weather conditions (e.g. temperature, wind speed) and output variables, external variables (e.g. temperature, wind speed) will be taken as important input variables, and in addition due to the important impact of the historical load demand of the home and the historical output power of the photovoltaic generation on the prediction results will also be taken as input variables.
Further supplementing the present solution, the two steps of data screening and preprocessing include checking all data to avoid data waste, and in addition, this step means using data for noise reduction, identifying trends and finding any important connections.
Further supplementing the technical scheme, dividing the data set in step three: the processed data set is divided into training, validation and test data sets.
The technical scheme is further supplemented by the following steps of selecting parameters of an ANN model: the calculations to find and mitigate complex correlations are the reason for using the parametric function.
In addition to the technical scheme, the number of the hidden layers and the number of the neurons are determined by a trial and error method;
(1) training function: Levenberg-Marquardt back-propagation.
(2) Transfer function: sigmoid function.
(3) The evaluation criterion is root mean square error.
(4) Stopping standard: once the error function has not been added, learning stops.
(5) Final prediction model evaluation: the predicted performance was evaluated by using the Mean Absolute Percent Error (MAPE).
(6) Input variables are: data analysis shows that the PV output power and the power load demand have large correlation with temperature and wind speed. Thus the weather conditions are taken as external variables of the model, where X1(t) is the temperature per hour, X2(t) is the wind speed per hour. Because the load demand of the previous hour and the load demand of the previous day are simultaneously in strong positive correlation with the current load demand, the two variables and the hour in the day are respectively used as the variable X3、X4、X5. The following external variables were used in the photovoltaic power prediction model: x1Temperature, X2Wind speed, X3Hour of the day, X4Previous hour data, X5Data of the same hour of the previous day. On the other hand, the following external variables are used for the home load prediction model: x1Temperature, X2Average load of the first two hours, X3Hour of the day, X4Data of the previous hour, X5Data of the same hour of the previous day.
(7) Number of hidden layers: 2 hidden layers;
(8) number of hidden neurons: each hidden layer has 10 neurons.
To further supplement the technical solution, in order to improve the performance of the prediction model and reduce the high error peak value to the maximum extent, external variables such as weather conditions are considered, and herein, the influence of the external variables such as the weather conditions on the ARIMAX and ANN prediction models is considered, and the prediction models are first divided into sub-models to evaluate the influence of the external variables on the animax and ARIMAX models, specifically as follows:
model a 1: ARIMAX has two external variables (X)1Temperature, X2-wind speed);
model a 2: ARIMAX has an external variable (X)1-temperature);
model a 3: ARIMAX has an external variable (X)2-wind speed);
model a 4: ARIMAX only has photovoltaic power generation and user load and does not have a model of an external variable;
model NN 1: the ANN model has several external variables (X)1Temperature, X2Wind speed, X3Hour of the day, X4Previous hour data, X5Data of the same hour of the previous day);
model NN 2: ANN models do not contain weather Condition variables (X)3Hour of the day, X4Previous hour data, X5-data of the same hour of the previous day);
model NN 3: ANN model has weather variables (X) only1Temperature, X2-wind speed).
For the test data, the overall performance of each sub-model short-term prediction model is shown in the following table:
TABLE 1 Overall Performance of short-term prediction model for each sub-model
Analysis Table 1 shows thatRIMAX predictive models when using ANN predictive models and taking into account external variables (X)1Temperature, X2Wind speed, X3Hour of the day, X4Previous hour data, X5Data of the same hour of the previous day), the prediction accuracy is higher.
The technical scheme is further supplemented, parameters of the ANN prediction model are optimized by adopting a GROM algorithm, and the method can be realized by the following steps:
the method comprises the following steps: taking some random learning parameters of the ANN prediction model as population initialization parameters, and calculating the average value of the population;
step two: evaluating the fitness of each model parameter by adopting a learning cost function in the ANN, then comparing the fitness of the population mean solution with the fitness of the worst solution, and replacing the worst solution with the population mean solution if the mean solution obtains a better fitness function value, wherein the process in the GROM aims to enhance the optimization speed to make the algorithm converge;
step three: creating a random solution vector in the group to determine and specify the direction of the next advance and the size of the motion; the optimal parameter solution is the solution with the minimum objective function value, and in the GROM, the parameter solution needs to be updated and moves towards the direction of the population optimal solution.
In general, the proposed GROM optimization technique does not have any adjustment steps for the optimization model, which helps to simplify the model, reduce convergence rate and computational cost. In the patent, the optimized model parameters are evaluated in a wide numerical range, and an optimal parameter solution is obtained, so that a model prediction result is obtained.
Further supplementing the present solution, the accuracy of the prediction model is determined by using the Mean Absolute Percentage Error (MAPE) and the Root Mean Square Error (RMSE), as shown in equations (4) and (5);
where l (t) is actual data, such as actual user load data;
is predicted data; t is the current time step, T is the total number of time steps (observations); MAPE and RMSE are the most commonly used evaluation methods for predictive models. MAPE is a scale-independent method that can be easily interpreted as a percentage, however if the actual data reading is zero, MAPE cannot be used because it would generate an undefined value. Therefore, RMSE is used herein to avoid this problem in predictive model evaluation. However, the evaluation methods such as RMSE and MAPE focus on the average value of the error and do not show the performance of the prediction model at each time step. For example, in some cases, the actual load demand and predicted load demand curves are of similar magnitude, but there is a time offset between the two profiles, so that extremely high error values may occur.
Further supplementing the technical scheme, the performance test of the prediction model adopts the following steps:
the method comprises the following steps: comparing the prediction model performance of different data profiles:
step two: the influence of the external variables on the predictive model is evaluated.
In the embodiment, the rated capacity of the inverter in the photovoltaic power generation system is 4kW, and the electricity consumption of each month in a year of a single user household is shown in the following table.
TABLE 1 Power consumption by Individual user households in a year and each month
| Month of the year | Electric power consumption-kWh | Month of the year | Electric power consumption- |
| 1 | 1089 | 7 | 1012 |
| 2 | 1080 | 8 | 1050 |
| 3 | 574 | 9 | 784 |
| 4 | 544 | 10 | 510 |
| 5 | 866 | 11 | 644 |
| 6 | 870 | 12 | 900 |
MAPE and RMSE (calculated daily during the test) for each model based on ANN-GROM model predictions and compared to conventional ANN models, ARIMAX model predictions, are shown in the following table.
TABLE 3 MAPE and RMSE for each model
MAPE and RMSE family load demand applications are calculated from the average of ten user loads without any significant deviation in these results. Generally, the mean-value method is one of the most common methods for solving the random problem. The ANN-GROM predictive model has the highest predictive accuracy for all data during testing, from an overall performance consideration.
First, predictive results are generated during testing based on three data sets for the conventional ANN and ARIMAX models and compared to actual data. For an actual user as an example, the net load curve and the prediction result curve in a day are shown in fig. 1 and fig. 2, respectively.
From the analysis in FIG. 1, the ARIMAX model predicted a significant peak at 8:00 was lost and the predicted value was significantly lower than the actual home user load. On the other hand, compared with the traditional artificial neural network and ANN-GROM model, the ARIMAX model has lower prediction result.
For all three types of data sets, analysis of Table 3 reveals that the ANN-GROM model predicts better performance than the conventional ANN and ARI MAX models. Compared with the traditional artificial neural network model, the MAPE of the ANN-GROM model is improved by 41.2%, 22.1% and 30.1% respectively for the data sets of the family user load, the photovoltaic output and the net load.
In addition, table 3 shows the highest RMSE values at which ARIMAX generated the predicted net demand curve. All predictive models exhibit lower predictive performance during prediction of net load demand compared to photovoltaic power generation and home consumer load prediction. This is mainly because the choice of external variables for both prediction techniques is chosen according to the correlation between weather conditions and both PV or domestic electrical loads, without taking into account the network load demand curve.
The smart meter data for 10 homes and photovoltaic systems are given in table 4 and used to predict the electricity load for an individual home.
Intelligent electric meter data for 410 households and photovoltaic power generation systems
The technical solutions described above only represent the preferred technical solutions of the present invention, and some possible modifications to some parts of the technical solutions by those skilled in the art all represent the principles of the present invention, and fall within the protection scope of the present invention.
Claims (10)
1. A low-voltage distribution network load prediction method considering distributed power supply influence is characterized by comprising the following working steps:
the method comprises the following steps: selecting input and output variables based on user load data of a low-voltage power grid comprising distributed photovoltaic;
step two: screening and preprocessing data, and identifying unusual observation variables and detection trends;
step three: dividing a data set into a training data set, a verification data set and a test data set;
step four: selecting ANN model parameters, and determining the number of hidden layers, the number of layers, a transfer function and an evaluation standard;
step five: training an ANN model;
step six: realizing an ANN model and outputting a result;
wherein, this technical scheme applies GROM in ANN model, and its implementation mode is based on Fibonacci function, namely
The series F0, F1, F2
Fn＝Fn-1+Fn-2 (1)
Where phi is called the golden ratio, for updating the search process and finding the optimal solution in two different stages, which comprises the following steps:
the first step is as follows: calculating the average value of all possible solutions for training the ANN network; then considering the fitness function, comparing the mean solution with the worst solution, and replacing the worst solution with the mean solution if the mean solution results in a better fitness function value, the purpose of this process is to speed up the algorithm and get it converged as soon as possible.
The second step is that: in order to determine the search direction, a random solution and a mean solution are selected to be compared to explore the influence of the random solution and the mean solution on the search motion, and the method can help to determine and optimize ANN model parameters and can avoid deviation of a prediction model caused by selection of additional parameters.
2. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence according to claim 1, wherein the output variables in the step one are as follows: predicting load demand at time t + i
Photovoltaic output power
Input variables are: due to the strong robustness between weather conditions (e.g. temperature, wind speed) and output variables, external variables (e.g. temperature, wind speed) will be taken as important input variables, and in addition due to the important impact of the historical load demand of the home and the historical output power of the photovoltaic generation on the prediction results will also be taken as input variables.
3. The method for forecasting the load of the low-voltage distribution network considering the distributed power supply influence is characterized in that the step two of data screening and preprocessing comprises checking all data to avoid data waste.
4. A method for load prediction of a low voltage distribution network taking into account the effects of distributed power sources according to claim 3, characterized by the fact that the data set is divided in three steps: the processed data set is divided into training, validation and test data sets.
5. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence according to claim 4, wherein the step of four ANN model parameter selection comprises the following steps: the calculations to find and mitigate complex correlations are the reason for using the parametric function.
6. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence according to claim 5, wherein the number of hidden layers and neurons is determined by a trial and error method.
7. The method as claimed in claim 6, wherein in order to improve the performance of the prediction model and minimize the high error peak, external variables such as weather conditions are considered, and the influence of the external variables such as weather conditions on the two prediction models, ARIMAX and ANN, respectively, is considered, and the prediction model is first divided into sub-models to evaluate the influence of the external variables on the models, animax and ARIMAX, respectively, as follows:
model a 1: ARIMAX has two external variables (X)1Temperature, X2-wind speed);
model a 2: ARIMAX has an external variable (X)1-temperature);
model a 3: ARIMAX has an external variable (X)2-wind speed);
model a 4: ARIMAX only has photovoltaic power generation and user load and does not have a model of an external variable;
model NN 1: the ANN model has several external variables (X)1Temperature, X2Wind speed, X3Hour of day, X4Data of the previous hour, X5Data of the same hour of the previous day);
model NN 2: ANN models do not contain weather Condition variables (X)3Hour of the day, X4Data of the previous hour, X5-data of the same hour of the previous day);
model NN 3: ANN model has weather variables (X) only1Temperature, X2-wind speed).
8. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence is characterized in that the GROM algorithm is adopted to optimize the parameters of the ANN prediction model, and the method can be realized by the following steps:
the method comprises the following steps: taking some random learning parameters of the ANN prediction model as population initialization parameters, and calculating the average value of the population;
step two: evaluating the fitness of each model parameter by adopting a learning cost function in the ANN, then comparing the fitness of a population mean solution with the fitness of a worst solution, and replacing the worst solution with a better fitness function value if the mean solution obtains the better fitness function value, wherein the process in the GROM aims to enhance the optimization speed so that the algorithm is converged;
step three: creating a random solution vector in the group to determine and specify the direction of the next advance and the size of the motion; the optimal parametric solution is the solution with the minimum objective function value, and in the GROM, the parametric solution needs to be updated and moved to the direction of the population optimal solution.
9. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence is characterized in that the accuracy of the prediction model is determined by using a Mean Absolute Percentage Error (MAPE) and a Root Mean Square Error (RMSE) as shown in the formulas (4) and (5);
where l (t) is actual data, such as actual user load data;
is predicted data; t is the current time step and T is the total number of time steps (observations).
10. The method for predicting the load of the low-voltage distribution network considering the distributed power supply influence is characterized in that the performance test of the prediction model is evaluated by adopting the following steps:
the method comprises the following steps: comparing the performance of the prediction models of different data profiles:
step two: the influence of the external variables on the predictive model is evaluated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210291236.XA CN114707713A (en) | 2022-03-23 | 2022-03-23 | Low-voltage distribution network load prediction method considering distributed power supply influence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210291236.XA CN114707713A (en) | 2022-03-23 | 2022-03-23 | Low-voltage distribution network load prediction method considering distributed power supply influence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114707713A true CN114707713A (en) | 2022-07-05 |
Family
ID=82169019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210291236.XA Pending CN114707713A (en) | 2022-03-23 | 2022-03-23 | Low-voltage distribution network load prediction method considering distributed power supply influence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707713A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115333100A (en) * | 2022-10-12 | 2022-11-11 | 四川中电启明星信息技术有限公司 | Roof photovoltaic power generation power cooperative control method and system |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205207A (en) * | 2021-04-19 | 2021-08-03 | 深圳供电局有限公司 | XGboost algorithm-based short-term power consumption load fluctuation prediction method and system |
-
2022
- 2022-03-23 CN CN202210291236.XA patent/CN114707713A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113205207A (en) * | 2021-04-19 | 2021-08-03 | 深圳供电局有限公司 | XGboost algorithm-based short-term power consumption load fluctuation prediction method and system |
Non-Patent Citations (3)
| Title |
|---|
| FERAS ALASALI等: "Forecasting and Modelling the Uncertainty of Low Voltage Network Demand and the Effect of Renewable Energy Sources", ENERGIES, vol. 14, no. 8, 12 April 2021 (2021-04-12), pages 2157 * |
| 吴雪莲 等: "基于BP神经网络-马尔科夫链的光伏发电预测", 电工电气, no. 03, 15 March 2014 (2014-03-15) * |
| 徐建中 等: "改进鲸鱼优化算法在电力负荷调度中的应用", 运筹与管理, no. 09, 25 September 2020 (2020-09-25) * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115333100A (en) * | 2022-10-12 | 2022-11-11 | 四川中电启明星信息技术有限公司 | Roof photovoltaic power generation power cooperative control method and system |
| CN115333100B (en) * | 2022-10-12 | 2022-12-16 | 四川中电启明星信息技术有限公司 | Roof photovoltaic power generation power cooperative control method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107516170B (en) | Difference self-healing control method based on equipment failure probability and power grid operation risk | |
| CN110909912B (en) | Park electric power system net load combination prediction method based on self-adaptive error feedback | |
| CN109861211B (en) | Dynamic reconfiguration method of power distribution network based on data driving | |
| CN110796307B (en) | Distributed load prediction method and system for comprehensive energy system | |
| CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| Wang et al. | Multi-objective energy management system for DC microgrids based on the maximum membership degree principle | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN111027775A (en) | Step hydropower station generating capacity prediction method based on long-term and short-term memory network | |
| CN112149890A (en) | Comprehensive energy load prediction method and system based on user energy label | |
| CN112232575A (en) | Comprehensive energy system regulation and control method and device based on multivariate load prediction | |
| Chang et al. | Feature extraction of non-intrusive load-monitoring system using genetic algorithm in smart meters | |
| Chen et al. | MultiCycleNet: multiple cycles self-boosted neural network for short-term electric household load forecasting | |
| CN114707713A (en) | Low-voltage distribution network load prediction method considering distributed power supply influence | |
| Poczeta et al. | Application of fuzzy cognitive maps to multi-step ahead prediction of electricity consumption | |
| CN117748444A (en) | Operation simulation method of power distribution system | |
| Li et al. | Residential Photovoltaic Power Forecasting Considering Division of Weather Type Index Interval | |
| CN114498753A (en) | Data-driven low-carbon ship micro-grid real-time energy management method | |
| Ming et al. | Electrical load prediction in energy internet via linear correlation coefficient approach | |
| CN113487062A (en) | Power load prediction method based on periodic automatic encoder | |
| Duong et al. | Application of Seasonal Trend Decomposition using Loess and Long Short-Term Memory in Peak Load Forecasting Model in Tien Giang | |
| Mellit et al. | Application of neural networks and genetic algorithms for predicting the optimal sizing coefficient of photovoltaic supply (PVS) systems | |
| Amreen et al. | Solar power forecasting using hybrid model | |
| Chinnadurrai et al. | Energy Management of a Microgrid based on LSTM Deep Learning Prediction Model | |
| Fu et al. | Electricity market price prediction based on differential evolutionary algorithm and neural network | |
| Fan et al. | Ultra-Short-Term Wind Power Forecasting Based on Fluctuation Pattern Clustering and Prediction |
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 |