CN115564128A - Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days - Google Patents
Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days Download PDFInfo
- Publication number
- CN115564128A CN115564128A CN202211285777.8A CN202211285777A CN115564128A CN 115564128 A CN115564128 A CN 115564128A CN 202211285777 A CN202211285777 A CN 202211285777A CN 115564128 A CN115564128 A CN 115564128A
- Authority
- CN
- China
- Prior art keywords
- day
- prediction
- load rate
- overload
- load
- 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 39
- 238000012937 correction Methods 0.000 title claims abstract description 31
- 238000012549 training Methods 0.000 claims description 15
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012986 modification Methods 0.000 claims description 2
- 230000004048 modification Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Mathematical Analysis (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Fuzzy Systems (AREA)
- Marketing (AREA)
- Pure & Applied Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Algebra (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of load prediction, and particularly relates to a distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days. According to the method, the time points with the heavy overload possibility are obtained by counting the probability of the heavy overload of the load rate at each time point in the near term, the time points with the heavy overload possibility are concentrated and corrected, and the correction efficiency is improved; in consideration of the influence of short-term external factors on the load rate peak value, the load rate similar day of the prediction day is found by means of the load rate curve of the previous day and the similar day of the prediction day, and then the initial predicted load rate curve is corrected at the moment when the overload probability exists according to the load rate peak value heights of the previous day and the similar day of the prediction day, so that the prediction accuracy of the overload part is improved, and the problem of overload missing judgment is solved.
Description
Technical Field
The invention belongs to the technical field of load prediction, and particularly relates to a distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days.
Background
In recent years, with the rapid development of social economy, the electricity consumption for production and living of people is remarkably increased, and particularly in the peak period of use of air conditioners in summer, the daily electricity consumption peaks in areas with intensive people and living and large industrial parks are relatively concentrated, so that heavy load and even overload of distribution and transformation are easily caused. Distribution transformer operating under heavy overload condition can cause transformer trouble, brings very big inconvenience for resident's life, influences production operating efficiency and causes huge economic loss. In order to avoid heavy overload of the distribution transformer and ensure stable operation of a distribution transformer line, it is important to perform risk early warning on the potential heavy overload distribution transformer.
At present, heavy overload early warning work is mainly carried out on the basis of predicting the load rate of distribution transformer, through a historical load rate data fluctuation rule and adding factors such as air temperature and the like which can directly or indirectly affect the load rate, a load rate prediction model is constructed by means of various machine learning methods, and finally, the heavy overload condition is judged on the basis of a load rate prediction result. In the field of machine learning, a time series method is applied to load rate prediction services more frequently, wherein an Autoregressive Integrated Moving Average model (ARIMA) is a mature algorithm, the ARIMA establishes a time series model of historical load rate fluctuation along with time, and the model predicts the future load rate through the learning of historical data. Also commonly used as a classical prediction method is a regression analysis method, which can learn the relationship between independent variables and dependent variables and establish a regression equation based on historical data. Time series methods and regression analysis methods are the mainstream techniques for load rate prediction. However, due to the influence of some short-term external factors, the prediction effect of the traditional machine learning model on the load rate peak value is difficult to stably and accurately achieve heavy overload early warning, so that a lot of heavy overload hidden dangers are missed to be judged; and the prediction result of the load rate is corrected by using a fixed coefficient, the pertinence to a heavy overload part is lacked, the later period needs manual experience, manpower and material resources are wasted, and the efficiency is low.
Disclosure of Invention
In order to overcome the problems in the prior art, the invention provides a distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days.
The technical scheme for solving the technical problems is as follows:
the distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days comprises the following steps:
step 1: acquiring relevant external influence data of a prediction day, a load rate x days before the prediction day and corresponding relevant external influence data as an initial data set;
step 2: preprocessing the initial data set to form a preprocessed data set;
and step 3: performing characteristic engineering operation on the preprocessed data set to form a training data set;
and 4, step 4: adopting LGBM to construct a load rate prediction model, and training the load rate prediction model by utilizing a training data set to obtain a trained load rate prediction model; obtaining a load prediction curve of a prediction day by using the trained load rate prediction model;
step 5, counting and predicting the probability of overload of the load rate at each moment point m days before the day, setting the overload probability threshold values of distribution and transformation lines of different types according to the overload probability, and acquiring the moment points with overload possibility;
step 6, acquiring load rate similar days of the prediction days;
and 7, correcting the initial predicted load rate curve at the moment when the overload probability exists according to the load rate peak heights of the day before the predicted day and the similar day.
Further, the relevant external influence data in step 1 include total active power, date attribute, weather attribute and temperature attribute.
Further, the preprocessing the initial data set in step 2 includes: abnormal data removing, missing data filling and classified data encoding.
Further, the feature engineering operation in step 3 specifically includes:
add date feature column: dividing the dates into 1 to 7 according to the week, adding feature columns from Monday to Sunday, adding holidays 1, adding holidays 0 for other dates, and highlighting the date types;
adding the total active characteristics of all the moments of the previous day;
adding temperature characteristics at each moment: adding 96 points of real-time temperature characteristics of a local line, and predicting the use of weather forecast temperature characteristics on the same day.
Further, the step of obtaining similar days in step 6 is: the load factor curve of n (n < m) days before the predicted day is taken, the day with the highest similarity to the load factor curve of the day before the predicted day is searched for in n-1 days excluding the day before the predicted day, and the day after the day is taken as the load factor similarity day of the predicted day.
Further, the load rate curve similarity calculation method comprises the following steps:
wherein E is i Relative error predicted for each time instant:
in the above formula, s is the result of the load rate of each time point on the similar day, t is the load rate of each time point on the predicted day, and N is the number of the predicted time points.
Further, the rule for modifying the initial predicted load factor curve in step 7 includes:
when the probability of occurrence of heavy overload is estimated to be =0 at a certain time point, and meanwhile, the prediction result of the load prediction curve is heavy overload, the load prediction curve is corrected to be in a normal state;
when the probability of occurrence of the overload is estimated to be =0 at a certain time point, and meanwhile, the prediction result of the load prediction curve is normal, no correction is carried out;
when the estimated occurrence probability of the overload is less than the overload probability threshold beta at a certain moment 0 or less, correcting the prediction result of the load prediction curve;
when the probability of occurrence of heavy overload is estimated to be more than or equal to a heavy overload probability threshold value beta at a certain time point, and the prediction result of the load prediction curve is in a normal state, taking the larger true value of the time point in the similar day and the day before the prediction day as a correction value;
when the probability of occurrence of heavy overload is estimated to be more than or equal to the heavy overload probability threshold beta at a certain time point, and the prediction result of the load prediction curve is heavy overload, no correction is carried out.
Compared with the prior art, the invention has the following technical effects:
according to the method, the time points with the heavy overload possibility are obtained by counting the probability of the heavy overload of the load rate at each time point in the near term, the time points with the heavy overload possibility are concentrated and corrected, and the correction efficiency is improved; in consideration of the influence of short-term external factors on the load rate peak value, the load rate similar day of the prediction day is found by means of the load rate curve of the previous day and the load rate similar day of the prediction day, and then the initial predicted load rate curve is corrected at the moment when the overload probability exists according to the load rate peak value heights of the previous day and the similar day of the prediction day, so that the prediction accuracy of the overload part is improved, and the problem of overload missing is solved.
Drawings
FIG. 1 is a schematic diagram of a process of correcting a load factor based on an LGBM model in combination with recent overload time probability statistics and similar days;
fig. 2 is a schematic diagram illustrating the probability statistics that heavy overload may occur at each moment of a distribution line 96 according to the present invention;
FIG. 3 is a comparison graph of the real value of the random distribution line in a certain day and the prediction result of the LGBM model;
fig. 4 is a comparison graph of the real value of the random distribution and transformation line in a certain day, the predicted value of the LGBM model and the result after correcting the predicted value.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
The invention aims at predicting a load rate value taking 15 minutes as a sampling interval (96 points) within 24 hours of a future day, and provides a distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days, which specifically comprises the following steps:
step 1: acquiring relevant external influence data of a prediction day, a load rate x days before the prediction day and corresponding relevant external influence data as an initial data set;
the relevant external influence data comprise total active power, date attributes, weather attributes and temperature attributes.
Step 2: preprocessing the initial data set to form a preprocessed data set;
because the abnormal data can reduce the short-term prediction effect of the linear regression model, in order to ensure the short-term prediction stability of the linear regression model, the data needs to be cleaned, the interference of abnormal values to the model learning process is reduced, the prediction precision is improved, the initial data set is preprocessed based on the abnormal data, and the data preprocessing step mainly comprises the following aspects:
and (3) abnormal data elimination: by setting a service rule, eliminating the load rate with huge fluctuation, checking by using historical data and the load rate value of a point near an abnormal point, and repairing the abnormal value while eliminating the abnormal value;
missing data filling: a small amount of missing data is supplemented by using a minimum dichotomy, measuring points of a large amount of missing data are correspondingly removed, the abnormal degree of the data is reduced, the availability of the data is improved, and null value filling is used for classified data;
classified data encoding: and (3) encoding the weather, holiday, day of week and other classification data, and converting the character string into a recognizable numerical type classification characteristic.
And step 3: performing characteristic engineering operation on the preprocessed data set to form a training data set;
the load rate is influenced by various external characteristic factors such as temperature and date attributes, other characteristics are considered to be added on the basis of the original characteristics, the characteristics in the initial data set can not improve the prediction capability of the model, part of useless characteristics are added, the prediction capability of the model is even reduced, and the training time is prolonged. And key features are selected from the initial data, so that the training efficiency of the model is improved. In order to solve the problems, the method carries out characteristic engineering operation on the preprocessed initial data, analyzes characteristic factors influencing the load rate and fluctuation, and screens important characteristics, so that the prediction accuracy of the model is improved.
The characteristic engineering part of the invention mainly processes the characteristics as follows:
add date feature column: the power utilization rules of different lines are influenced by date types, in order to strengthen the model to learn the relationship between different types of dates and load rates, the dates are divided into 1 to 7 according to the week, a characteristic column from Monday to Sunday is added, holidays are 1, other dates are 0, a holiday characteristic column is added, and the date types are highlighted.
Adding the total active characteristics of the previous day at each moment: the real-time change trend of the total active power and the change trend of the load rate have certain relevance, and the total active power of the day of the forecast day is unknown, so that the total active power of the previous day is input.
Adding temperature characteristics at each moment: the change of real-time temperature can cause the load factor to increase suddenly and decrease suddenly, and during the high temperature period in summer, a user can use the air conditioner in a centralized way in the temperature peak value interval of one day, so that the load factor can be increased to overload in a short time, and 96-point real-time temperature characteristics of a local line are added to predict the use of weather forecast temperature characteristics on the same day.
Data and features determine the upper limit of machine learning, and models and algorithms only approximate this upper limit. Therefore, the characteristic engineering has very important significance for improving the training effect of the model and enhancing the performance of the model.
And 4, step 4: adopting LGBM to construct a load rate prediction model, and training the load rate prediction model by utilizing a training data set to obtain a trained load rate prediction model; and obtaining a load prediction curve of a prediction day by using the trained load rate prediction model.
Compared with the GBDT algorithm, the LGBM has the advantages of parallel training, high training efficiency, low memory consumption and capability of accurately and efficiently processing mass data.
The basic principle of the LGBM comes from a histogram algorithm, which can construct a histogram for each feature, and convert the traversal samples into the traversal histogram, and this process does not require additional memory space, thus greatly reducing the time complexity. And (leaf-wise) decision tree growth strategy adopted by the LGBM, which traverses all leaves, locates the leaf with the maximum splitting gain, splits the leaf and repeats the process. This strategy reduces many errors, but is prone to overfitting phenomena. The LGBM uses a unilateral gradient algorithm in the training process, and the method can filter samples with small gradient, thereby reducing the calculation amount of the model. However, the LGBM is sensitive to noise, and the constructed histogram is coarse, and some accuracy is lost.
The prediction of the 96-point load rate can be regarded as a prediction problem of a time series, machine learning is a mainstream method for solving the time series, wherein the LGBM model has a prominent effect and can convert a time series problem into supervised learning, and the LGBM model can solve most time series prediction models, support complex data modeling, support multivariate cooperative regression and support a nonlinear problem.
Step 5, counting and predicting the probability of overload of the load rate at each moment point m days before the day, setting the overload probability threshold values of distribution and transformation lines of different types according to the overload probability, and acquiring the moment points with overload possibility;
the daily load rate curves of different types of distribution transformer lines are greatly different, the time periods of peak power loads are different, but the time periods of peak load rates of the same distribution transformer line are approximately similar.
And (3) taking the load rate of 96 points per day m days before the forecast date, giving a heavy load warning because the load rate is more than or equal to 80 and less than 100, giving an overload warning because the load rate is more than or equal to 100, giving a heavy overload early warning at the point if the load rate at a certain moment is more than or equal to 78, calculating according to the heavy overload condition by default, and counting the probability of heavy overload at each moment point of 96 points in m days. The overload probability in m days is statistically analyzed, so that overload probability threshold values beta exist in different distribution lines, when the overload probability is higher than the overload probability threshold value beta, the overload hidden danger at the moment is large, and the overload probability threshold value beta can be set according to statistics and experience.
Step 6, searching for a similar day of the load rate of the forecast day;
because the peak height of the daily load rate is greatly influenced by various recent external burst factors, the peak height of the daily load rate similar to that in the recent n (n < m) days is used as a reference.
And taking n days before the predicted day, searching the day with the highest similarity to the load rate curve of the day before the predicted day in n-1 days without the day before the predicted day according to a load rate curve similarity formula, and taking the day after the day as the load rate similar day of the predicted day.
Assuming that the load rate result of each time point on the similar day is s, the load rate of each time point on the predicted day is t, the similarity is sim, and the load rate curve similarity calculation formula of each time point is as follows:
in the above formula, E i The relative error of prediction for each time point, N is the number of predicted time pointsIf the calculation is carried out according to the time point of 15 minutes, N is taken as 96.
And 7, correcting the initial predicted load rate curve at the moment when the overload probability exists according to the load rate peak heights of the day before the predicted day and the similar day.
The accuracy of load rate peak prediction directly affects the heavy overload decision accuracy. The short-term load rate prediction is greatly influenced by various recent external burst factors (such as weather, epidemic situation and the like), and the uncertainty of the short-term prediction of the LGBM model is increased due to the external burst conditions, so that the LGBM model is difficult to learn the change rule of a load rate peak value area. Therefore, there is a need to increase the certainty of short-term load rate predictions based on recent load rate curve changes, and the present invention modifies the initial predicted load rate curve at points in time when there is a heavy overload probability based on the load rate peak heights on the day before and on similar days of the prediction.
The statistical analysis of the overload probability in m days shows that when the overload probability is higher than the overload probability threshold value beta, the overload hidden danger at the moment is larger. Thus, 96 points can be classified into three types, respectively: the heavy overload probability =0,0 is not less than the heavy overload probability < beta, and the heavy overload probability is not less than beta.
When the probability =0 of heavy overload at a certain time point and the LGBM prediction result is heavy overload, the LGBM is corrected to be in a normal state (below 80); when the probability =0 of the occurrence of the heavy overload at a certain time point and the LGBM prediction result is normal, not correcting; when the probability of occurrence of heavy overload is less than beta at a certain moment 0 or less, the LGBM prediction result is not corrected; when the probability of overload at a certain time point is more than or equal to beta and the LGBM prediction result is in a normal state, taking the larger true value of the time point in the similar day and the day before the prediction day as a correction value; and when the probability of overload at a certain time point is more than or equal to beta, and the LGBM forecasting result is overload, no correction is carried out.
According to the shape of the load rate curve of the day before the forecast day, the date with the highest load rate curve similarity is found from the near future, and the next day is taken as the similar day of the forecast day load rate curve; then, comparing the load rate of the day before the forecast day and the similar day of the forecast day at each moment, and reserving a value which is easier to trigger heavy overload; and finally, correcting the original prediction curve under a correction rule based on the heavy overload probability statistics. Compared with the output result of the LGBM model, the corrected load rate curve can better solve the situation of missed judgment of the future potential heavy overload, the correction method can reduce manual intervention and improve the accuracy of heavy overload judgment.
Effect of the experiment
The method for judging the heavy overload precision of 96 points in one day in the experiment of the invention comprises the following steps: judging whether the load rate true value and the predicted value of 96 points on the predicted day are heavy overload point by point, wherein the overall prediction precision of the distribution and transformation line in one day is PR total The calculation formula is as follows:
wherein N is correct To predict the correct number of spots, N total Is the total number of points (N) total ＝96)。
In order to verify the feasibility and the usability of the invention, a large number of experiments are carried out, and partial experiment results are as follows:
(I) statistical analysis of probability of heavy overload occurrence at different lines at different recent time points
The result of randomly counting the probability of heavy overload occurring at each time point of the near 96 points of a distribution transformer line is shown in fig. 2, and it can be seen that the time period of heavy overload occurring each day is approximately from 0 point to 4 points 30 minutes and from 18 points 30 to 23 points 45 minutes, the probability of heavy overload occurring at night is far greater than that in the early morning, and the probability of heavy overload occurring in the time period from 4 points 30 to 18 points 30 minutes is zero, so that in the work of judging whether heavy overload occurs in the distribution transformer, only the time point with the hidden danger of heavy overload needs to be corrected.
(II) comparing the prediction result of LGBM model with the true value
A line is randomly selected, dates are randomly screened, load rate prediction is carried out on the day of 6 months and 24 days in 2022, the result is shown in fig. 3, wherein a dotted line represents an initial load rate prediction curve output by the LGBM model, a solid line represents a real load rate curve on the day of prediction, and it can be obviously seen that the curve output by simply using the LGBM model is relatively similar in overall shape, and particularly, the similarity is higher in a part with a low intermediate load rate. However, in the graph, the load rates of two parts at two ends in the morning and at night are high, the overload is caused at some time points, the prediction result of the LGBM model to the part is much lower than the real result, and the overload hidden danger cannot be timely and accurately pre-warned in the work process.
(III) comparing the corrected prediction result with the prediction result and the true value of the LGBM model
As shown in fig. 4, the dashed line represents the load rate prediction curve modified by the method of the present invention, the solid line represents the real load rate curve, and the dashed dotted line represents the initial prediction curve output by the LGBM, it can be seen that the modified curve is closer to the real curve at the portion with higher load rate at the two ends, and the heavy overload hidden danger can be found in time.
The accuracy of the different distribution line overload correction prediction results and the initial prediction results is shown in the tables 1 and 2.
Table 1 comparison table of overload correction prediction results of distribution transformer in different distribution transformer lines at 2022 year, 6 month and 26 day
Table 2 comparison of overload correction prediction results of distribution transformer load in 2022, 7 months, 24 days for different distribution transformer lines
Tables 1 and 2 represent a comparison of results of random 10 different types of lines versus LGBM model predictions and post-correction predictions for 26 days at 6 months and 24 days at 7 months at 2022, respectively. According to research, the average prediction accuracy of distribution transformer overload is improved on 10 lines of different types. Compared with the manual correction based on subjective experience, the correction method is more intelligent and automatic.
The experimental results show that the short-term distribution transformer load rate correction prediction method based on the heavy overload probability statistics and the similar days can effectively improve the accuracy of short-term distribution transformer heavy overload prediction.
Because the probability that each distribution and transformation line is in a heavy overload state at 96 moments in a day is low, the accuracy rate of the overall load rate curve is high due to the fact that the ratio difference between heavy overload and non-heavy overload is too large, but the accuracy rate of predicting and judging whether the heavy overload state exists is low. And actually, in the LGBM model heavy overload prediction process, the prediction effect on the heavy overload part is poor.
Therefore, the invention provides a short-term load rate correction and prediction method, which is used for counting the probability of heavy overload of the load rate at each time point in the near term, acquiring the time point with the possibility of heavy overload, focusing on correcting the time point with the possibility of heavy overload and improving the correction efficiency; in consideration of the influence of short-term external factors on the load rate peak value, the load rate similar day of the prediction day is found by means of the load rate curve of the previous day and the similar day of the prediction day, and then the initial predicted load rate curve is corrected at the moment when the overload probability exists according to the load rate peak value heights of the previous day and the similar day of the prediction day, so that the prediction accuracy of the overload part is improved, and the problem of overload missing judgment is solved.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (7)
1. The distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days is characterized by comprising the following steps of:
step 1: acquiring relevant external influence data of a prediction day, a load rate x days before the prediction day and corresponding relevant external influence data as an initial data set;
step 2: preprocessing the initial data set to form a preprocessed data set;
and step 3: performing characteristic engineering operation on the preprocessed data set to form a training data set;
and 4, step 4: adopting LGBM to construct a load rate prediction model, and training the load rate prediction model by utilizing a training data set to obtain a trained load rate prediction model; obtaining a load prediction curve of a prediction day by using the trained load rate prediction model;
step 5, counting and predicting the probability of overload of the load rate at each moment point m days before the day, setting the overload probability threshold values of distribution and transformation lines of different types according to the overload probability, and acquiring the moment points with overload possibility;
step 6, acquiring load rate similar days of the forecast days;
and 7, correcting the initial predicted load rate curve at the moment when the overload probability exists according to the load rate peak heights of the day before the predicted day and the similar day.
2. The distribution transformer load rate modification prediction method based on heavy overload probability statistics and similar days as claimed in claim 1, wherein the relevant external influence data in step 1 comprise total active power, date attribute, weather attribute and temperature attribute.
3. The method according to claim 1, wherein the preprocessing the initial data set in step 2 comprises: abnormal data removing, missing data filling and classified data encoding.
4. The distribution transformer load rate correction and prediction method based on overload probability statistics and similar days as claimed in claim 1, wherein the characteristic engineering operation in step 3 specifically comprises:
add date feature column: dividing the dates into 1 to 7 according to the week, adding feature columns from Monday to Sunday, adding holidays 1, adding holidays 0 for other dates, and highlighting the date types;
adding the total active characteristics of all the moments of the previous day;
adding temperature characteristics at each moment: adding 96 points of real-time temperature characteristics of a local line, and predicting the use of weather forecast temperature characteristics on the same day.
5. The method for correcting and predicting the distribution transformer load rate based on the heavy overload probability statistics and the similar days according to claim 1, wherein the step of acquiring the similar days in the step 6 comprises the following steps: a load factor curve of n (n < m) days before the predicted day is obtained, a day with the highest similarity to the load factor curve of the day before the predicted day is searched for in n-1 days excluding the day before the predicted day, and the day after the day is taken as the load factor similar day of the predicted day.
6. The distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days according to claim 5, wherein in the step 6, a day with the highest similarity to a load rate curve of a day before the predicted day is searched for in n-1 days without the day before the predicted day, and the load rate curve similarity is adopted for calculation, specifically:
wherein E is i Relative error predicted for each time instant:
in the above formula, s is the result of the load rate of each time point on the similar day, t is the load rate of each time point on the predicted day, and N is the number of the predicted time points.
7. The method according to claim 5, wherein the modifying the initial predicted load rate curve in step 7 comprises:
when the probability of occurrence of heavy overload is estimated to be =0 at a certain time point, and meanwhile, the prediction result of the load prediction curve is heavy overload, the load prediction curve is corrected to be in a normal state;
when the probability of occurrence of the overload is estimated to be =0 at a certain time point, and meanwhile, the prediction result of the load prediction curve is normal, no correction is carried out;
when the estimated occurrence probability of the overload is less than the overload probability threshold beta at a certain moment 0 or less, correcting the prediction result of the load prediction curve;
when the estimated occurrence probability of the overload at a certain time point is larger than or equal to the overload probability threshold beta, and the prediction result of the load prediction curve is in a normal state, taking the larger true value of the time point in the similar day and the day before the prediction day as a correction value;
when the probability of occurrence of heavy overload is estimated to be more than or equal to the heavy overload probability threshold value beta at a certain time point, and meanwhile, the prediction result of the load prediction curve is heavy overload, no correction is carried out.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211285777.8A CN115564128A (en) | 2022-10-20 | 2022-10-20 | Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211285777.8A CN115564128A (en) | 2022-10-20 | 2022-10-20 | Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115564128A true CN115564128A (en) | 2023-01-03 |
Family
ID=84746759
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211285777.8A Pending CN115564128A (en) | 2022-10-20 | 2022-10-20 | Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115564128A (en) |
-
2022
- 2022-10-20 CN CN202211285777.8A patent/CN115564128A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105069525B (en) | Round-the-clock 96 Day Load Curve Forecastings and optimization update the system | |
| CN110503256B (en) | Short-term load prediction method and system based on big data technology | |
| CN111191193A (en) | Long-term soil temperature and humidity high-precision prediction method based on autoregressive moving average model | |
| CN110909958A (en) | Short-term load prediction method considering photovoltaic grid-connected power | |
| CN110852496A (en) | Natural gas load prediction method based on LSTM recurrent neural network | |
| CN111738477A (en) | Deep feature combination-based power grid new energy consumption capability prediction method | |
| CN112149902A (en) | Subway short-time arrival passenger flow prediction method based on passenger flow characteristic analysis | |
| CN116227637A (en) | Active power distribution network oriented refined load prediction method and system | |
| CN115860797A (en) | Electric quantity demand prediction method suitable for new electricity price reform situation | |
| CN112801388B (en) | Power load prediction method and system based on nonlinear time series algorithm | |
| CN116826745B (en) | Layered and partitioned short-term load prediction method and system in power system background | |
| CN117277312A (en) | Gray correlation analysis-based power load influence factor method and equipment | |
| CN115564128A (en) | Distribution transformer load rate correction prediction method based on heavy overload probability statistics and similar days | |
| CN113837486B (en) | RNN-RBM-based distribution network feeder long-term load prediction method | |
| CN114676931B (en) | Electric quantity prediction system based on data center technology | |
| CN113449933B (en) | Regional medium-term load prediction method and device based on clustering electric quantity curve decomposition | |
| CN114997470A (en) | Short-term power load prediction method based on LSTM neural network | |
| CN116777027A (en) | Load prediction method and system for abnormal days | |
| CN114529071A (en) | Method for predicting power consumption of transformer area | |
| CN110175705B (en) | Load prediction method and memory and system comprising same | |
| CN112801333B (en) | XGBoost-based power distribution network line summer peak load prediction method | |
| CN117291299B (en) | Moon electricity quantity prediction method considering various influence factors | |
| CN117910625A (en) | Gas load prediction method and system | |
| CN116228285A (en) | Chinese medicinal material price index research method based on big data analysis | |
| CN117151266A (en) | XGBoost-based supply chain performance risk prediction method |
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 |