CN116701869A - Peak regulation demand prediction method and terminal for power system - Google Patents
Peak regulation demand prediction method and terminal for power system Download PDFInfo
- Publication number
- CN116701869A CN116701869A CN202310674840.5A CN202310674840A CN116701869A CN 116701869 A CN116701869 A CN 116701869A CN 202310674840 A CN202310674840 A CN 202310674840A CN 116701869 A CN116701869 A CN 116701869A
- Authority
- CN
- China
- Prior art keywords
- peak shaving
- data
- value
- demand
- shaving demand
- 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 29
- 238000013528 artificial neural network Methods 0.000 claims abstract description 37
- 230000006870 function Effects 0.000 claims abstract description 33
- 230000015654 memory Effects 0.000 claims abstract description 29
- 238000010219 correlation analysis Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 28
- 230000007787 long-term memory Effects 0.000 claims description 16
- 238000012216 screening Methods 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 5
- 230000006403 short-term memory Effects 0.000 claims description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000009434 installation Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or 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
-
- 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
-
- 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)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a peak shaving demand prediction method and a terminal of an electric power system, which are characterized in that correlation analysis is carried out on characteristic quantity data and peak shaving demand historical data which influence peak shaving demands to obtain correlation coefficients among different data, input variables of a model are determined from the characteristic quantity data on the basis of the correlation coefficients among different data, the peak shaving demand historical data are determined to be output variables of the model, long-period memory neural network prediction models are trained by using the characteristic quantity data and the peak shaving demand historical data, the prediction precision of the model is effectively improved, and finally probability density function fitting is carried out on deviation values between peak shaving demand predicted values and peak shaving demand true values which are output by the trained prediction model, so that a peak shaving demand prediction interval which meets the confidence level requirements is obtained, probability distribution characteristics of prediction deviation can be fully considered, and the adaptability of a prediction result is better, and the accuracy and the adaptability of peak shaving demand prediction are improved.
Description
Technical Field
The invention relates to the technical field of power grid demand prediction, in particular to a peak shaving demand prediction method and a terminal of a power system.
Background
In recent years, the installed capacity and the power generation amount ratio of renewable energy sources are continuously improved, the renewable energy sources represented by wind power and photovoltaic have the characteristics of output fluctuation and uncertainty, and after the change of the power demand of a user at the load side is overlapped, the net load curve fluctuation of a power system is further increased. In addition, the rapid increase of the installation scale of renewable energy sources also reduces the proportion of the installation scale of conventional machine to a certain extent, so that the traditional peak shaving resources are reduced, the peak shaving pressure of a power grid is further increased, and how to accurately predict the peak shaving requirement of a power system in a high-proportion renewable energy source installation scene is one of the key technical problems to be solved.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the peak shaving demand prediction method and the terminal for the power system can improve accuracy and adaptability of peak shaving demand prediction.
In order to solve the technical problems, the invention adopts a technical scheme that:
a peak shaving demand prediction method of an electric power system comprises the following steps:
acquiring peak shaving demand historical data and characteristic quantity data influencing peak shaving demands of an electric power system, and carrying out correlation analysis based on the characteristic quantity data and the peak shaving demand historical data to obtain correlation coefficients among different data;
determining an input variable of a long-term memory neural network prediction model from the characteristic quantity data based on the correlation coefficient among different data, and determining the peak shaving demand historical data as an output variable of the long-term memory neural network prediction model;
training the long-period and short-period memory neural network prediction model by using the input variable and the output variable to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on the input variable;
and determining a peak shaving demand real value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data, and performing probability density function fitting on a deviation value between the peak shaving demand predicted value and the peak shaving demand real value to obtain a peak shaving demand predicted interval meeting the confidence level requirement.
In order to solve the technical problems, the invention adopts another technical scheme that:
a peak shaving demand prediction terminal for an electrical power system, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring peak shaving demand historical data and characteristic quantity data influencing peak shaving demands of an electric power system, and carrying out correlation analysis based on the characteristic quantity data and the peak shaving demand historical data to obtain correlation coefficients among different data;
determining an input variable of a long-term memory neural network prediction model from the characteristic quantity data based on the correlation coefficient among different data, and determining the peak shaving demand historical data as an output variable of the long-term memory neural network prediction model;
training the long-period and short-period memory neural network prediction model by using the input variable and the output variable to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on the input variable;
and determining a peak shaving demand real value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data, and performing probability density function fitting on a deviation value between the peak shaving demand predicted value and the peak shaving demand real value to obtain a peak shaving demand predicted interval meeting the confidence level requirement.
The invention has the beneficial effects that: the correlation analysis is carried out on the basis of the characteristic quantity data influencing the peak shaving demand and the peak shaving demand historical data to obtain correlation coefficients among different data, the input variable of the model is determined from the characteristic quantity data on the basis of the correlation coefficients among different data, the peak shaving demand historical data is determined to be the output variable of the model, the long-period memory neural network prediction model and the short-period memory neural network prediction model are used for training, the prediction precision of the model is effectively improved, finally probability density function fitting is carried out on the deviation value between the peak shaving demand predicted value and the peak shaving demand true value output by the trained prediction model, the peak shaving demand predicted interval meeting the confidence level requirement is obtained, the probability distribution characteristics of the prediction deviation can be fully considered, the adaptability of the predicted result is better, and the accuracy and the adaptability of the peak shaving demand prediction are improved.
Drawings
FIG. 1 is a flowchart illustrating a method for predicting peak shaving demand of an electric power system according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a peak shaving demand prediction terminal of an electric power system according to an embodiment of the present invention;
fig. 3 is a schematic diagram of calculation results of different correlation coefficients in the power system peak shaving demand prediction method according to an embodiment of the present invention;
fig. 4 is a comparison chart between a peak shaving demand prediction value based on a peak shaving demand prediction model of the power system and a peak shaving demand prediction interval meeting a confidence level requirement in the peak shaving demand prediction method of the power system according to the embodiment of the present invention.
Detailed Description
In order to describe the technical contents, the achieved objects and effects of the present invention in detail, the following description will be made with reference to the embodiments in conjunction with the accompanying drawings.
Referring to fig. 1, an embodiment of the present invention provides a power system peak shaving demand prediction method, including the steps of:
acquiring peak shaving demand historical data and characteristic quantity data influencing peak shaving demands of an electric power system, and carrying out correlation analysis based on the characteristic quantity data and the peak shaving demand historical data to obtain correlation coefficients among different data;
determining an input variable of a long-term memory neural network prediction model from the characteristic quantity data based on the correlation coefficient among different data, and determining the peak shaving demand historical data as an output variable of the long-term memory neural network prediction model;
training the long-period and short-period memory neural network prediction model by using the input variable and the output variable to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on the input variable;
and determining a peak shaving demand real value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data, and performing probability density function fitting on a deviation value between the peak shaving demand predicted value and the peak shaving demand real value to obtain a peak shaving demand predicted interval meeting the confidence level requirement.
From the above description, the beneficial effects of the invention are as follows: the correlation analysis is carried out on the basis of the characteristic quantity data influencing the peak shaving demand and the peak shaving demand historical data to obtain correlation coefficients among different data, the input variable of the model is determined from the characteristic quantity data on the basis of the correlation coefficients among different data, the peak shaving demand historical data is determined to be the output variable of the model, the long-period memory neural network prediction model and the short-period memory neural network prediction model are used for training, the prediction precision of the model is effectively improved, finally probability density function fitting is carried out on the deviation value between the peak shaving demand predicted value and the peak shaving demand true value output by the trained prediction model, the peak shaving demand predicted interval meeting the confidence level requirement is obtained, the probability distribution characteristics of the prediction deviation can be fully considered, the adaptability of the predicted result is better, and the accuracy and the adaptability of the peak shaving demand prediction are improved.
Further, performing correlation analysis based on the feature quantity data and the peak shaving demand historical data to obtain correlation coefficients between different data includes:
and respectively calculating a first correlation coefficient between each characteristic quantity data and the peak shaving demand historical data and a second correlation coefficient between different characteristic quantity data by using a Pearson correlation coefficient formula.
From the above description, the first correlation coefficient can represent the influence degree of each feature quantity data on the peak shaving requirement, the second correlation coefficient can represent the mutual influence degree of different feature quantity data, and the first correlation coefficient and the second correlation coefficient are beneficial to the subsequent screening of the feature quantity data most relevant to the peak shaving requirement to train the prediction model, so that the accuracy of the prediction model is improved.
Further, the determining the input variable of the long-term memory neural network prediction model from the feature quantity data based on the correlation coefficient between the different data includes:
judging whether the absolute value of the first correlation coefficient is smaller than or equal to a first preset value, if not, determining the characteristic quantity data as the characteristic quantity data after preliminary screening, and if so, eliminating the characteristic quantity data corresponding to the first correlation coefficient from the characteristic quantity data to obtain the characteristic quantity data after preliminary screening;
judging whether the absolute value of the first correlation coefficient in the preliminarily screened feature quantity data is larger than a first preset value and smaller than a second preset value, if not, determining the preliminarily screened feature quantity data as final feature quantity data, if so, determining the feature quantity data corresponding to the first correlation coefficient as target feature quantity data, judging whether the absolute value of the second correlation coefficient of the target feature quantity data is larger than or equal to a third preset value, if so, comparing the first correlation coefficient of the target feature quantity data, and eliminating the feature quantity data with the minimum absolute value of the first correlation coefficient to obtain final feature quantity data;
and determining the final characteristic quantity data as input variables of a long-term and short-term memory neural network prediction model.
From the description, the data with the correlation coefficient not conforming to the condition is removed from the characteristic quantity data according to the correlation coefficient among different data, the retained characteristic quantity data is the data which can influence peak shaving requirements most, and the data is used as the input variable of the long-short-term memory neural network prediction model, so that the accuracy of the prediction model is improved effectively.
Further, the training the long-short-term memory neural network prediction model by using the input variable and the output variable to obtain a peak shaving demand prediction model of the power system includes:
dividing the input variable and the output variable into four-quarter data according to quarters respectively;
randomly extracting one fifth combination from each quarter of the four-quarter data of the input variable and each quarter of the four-quarter data of the output variable to form a test set;
combining the remaining four fifths of the four quarters of the input variable and the four quarters of the output variable to form a training set;
training the long-short-period memory neural network prediction model by using the test set and the training set to obtain a peak regulation demand prediction model of the power system;
the obtaining a peak shaving demand prediction value based on the input variable using the power system peak shaving demand prediction model includes:
and predicting the input variables in the training set or the test set by using the power system peak shaving demand prediction model to obtain a peak shaving demand prediction value.
From the above description, after the input variable and the output variable are divided into four-quarter data according to the quarters, a test set and a training set are randomly formed to train the long-term and short-term memory neural network prediction model, so that a better training effect can be ensured.
Further, the characteristic quantity data affecting the peak shaving demand includes a daily load maximum value;
and performing probability density function fitting on the deviation value between the peak shaving demand predicted value and the peak shaving demand true value to obtain a peak shaving demand predicted interval meeting the confidence level requirement, wherein the peak shaving demand predicted interval comprises the following steps:
performing per unit processing on the peak shaving demand predicted value and the peak shaving demand real value by taking the daily load maximum value as a reference to obtain the processed peak shaving demand predicted value and peak shaving demand real value;
calculating a predicted deviation value between the processed peak shaving demand predicted value and the peak shaving demand true value;
performing fitting calculation on the predicted deviation value by using non-parameter kernel density estimation to obtain a probability density function of the predicted deviation value;
determining a confidence level of the predicted deviation value;
and determining a peak shaving demand prediction interval meeting the confidence level requirement based on the peak shaving demand prediction value, the confidence level and the probability density function.
Further, the performing a fitting calculation on the predicted deviation value by using the non-parameter kernel density estimation, and obtaining a probability density function of the predicted deviation value includes:
wherein f h A probability density function representing the predicted deviation value, N representing the number of data of the predicted deviation value, h representingBandwidth, K (·) represents the kernel function, x represents the offset argument, x i Representing the ith sample in the predicted deviation value.
From the above description, the non-parameter kernel density estimation is used to perform fitting calculation on the predicted deviation value, so as to obtain a probability density function of the predicted deviation value, and the peak shaving demand prediction interval meeting the confidence level requirement is determined based on the peak shaving demand predicted value, the confidence level and the probability density function, so that the probability distribution characteristics of the predicted deviation are fully considered, and the adaptability of the predicted result is effectively improved.
Further, the obtaining peak shaving demand history data and feature quantity data affecting the peak shaving demand of the power system includes:
acquiring peak shaving demand historical data of an electric power system and historical data which corresponds to the peak shaving demand historical data and affects peak shaving demands;
and extracting characteristic quantity data affecting the peak shaving demand from the historical data affecting the peak shaving demand by taking the day as a unit.
From the above description, it is known that extracting feature quantity data affecting peak shaving demands from historical data affecting peak shaving demands in units of days can improve the overall processing efficiency of the algorithm and the effectiveness of the algorithm.
Further, the historical data affecting peak shaving requirements comprise historical data of loads, wind power output and photovoltaic output;
the characteristic quantity data influencing peak shaving demands comprise a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value, a photovoltaic output maximum value and a correlation coefficient of load and wind and light total output;
the extracting, from the historical data affecting the peak shaving demand in units of days, feature quantity data affecting the peak shaving demand includes:
respectively counting historical data of the load, the wind power output and the photovoltaic output by taking a day as a unit to obtain a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value and a photovoltaic output maximum value;
and calculating according to historical data of the load, wind power output and photovoltaic output by using a pearson correlation coefficient formula by taking a day as a unit to obtain a correlation coefficient of the load and the total wind and light output.
From the above description, it can be seen that by extracting the correlation coefficients of the daily load maximum value, the daily load minimum value, the wind power output maximum value, the wind power output minimum value, the photovoltaic output maximum value and the total wind and light output, the prediction model can be more effectively trained by using the characteristic quantity data affecting the peak shaving demand, so as to realize the peak shaving demand prediction of the power system.
Further, the pearson correlation coefficient formula is:
wherein ρ is HK Represents the correlation coefficient between the data series H and the data series K, n represents the data number of the data series, H i Represents the ith data, K, in the data series H i Represents the ith data in the data series K,data mean value representing data series H, +.>The data average of data series K is shown.
The above description shows that the pearson correlation coefficient is simple to calculate, and can better reflect the relation between two data, so as to conveniently screen out the feature quantity data which is favorable for model training.
Referring to fig. 2, another embodiment of the present invention provides a peak shaving demand prediction terminal for an electric power system, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor implements each step in the peak shaving demand prediction method for the electric power system when executing the computer program.
The peak shaving demand prediction method and the terminal for the power system can be applied to the power system in the high-proportion renewable energy installation scene, and the following description is given by a specific embodiment:
example 1
Referring to fig. 1, 3 and 4, a peak shaving demand prediction method for an electric power system according to the present embodiment includes the steps of:
s1, acquiring peak shaving demand historical data and characteristic quantity data affecting peak shaving demands of an electric power system, and carrying out correlation analysis based on the characteristic quantity data and the peak shaving demand historical data to obtain correlation coefficients among different data, wherein the method specifically comprises the following steps of:
the historical data affecting peak shaving demands comprise historical data of loads, wind power output and photovoltaic output; the characteristic quantity data influencing peak regulation requirements comprise a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value, a photovoltaic output maximum value and a correlation coefficient of load and wind and light total output, and the characteristic quantity data units except the correlation coefficient of the load and the wind and light total output are MW.
S11, acquiring peak shaving demand historical data of the power system and historical data which corresponds to the peak shaving demand historical data and affects peak shaving demands;
specifically, peak shaving demand historical data of a power system with a time span of one year and historical data corresponding to the peak shaving demand historical data and affecting peak shaving demands are obtained, the peak shaving demand historical data is determined by daily load curve change, system rotation standby demand, wind power, photovoltaic power output and other factors, and 365 daily peak shaving demand historical data are obtained in total in one year.
S12, extracting characteristic quantity data affecting the peak shaving demand from the historical data affecting the peak shaving demand by taking a day as a unit, wherein the characteristic quantity data specifically comprises the following steps:
s121, respectively counting historical data of the load, the wind power output and the photovoltaic output by taking a day as a unit to obtain a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value and a photovoltaic output maximum value;
s122, calculating according to historical data of the load, wind power output and photovoltaic output by using a pearson correlation coefficient formula by taking a day as a unit to obtain a correlation coefficient of the load and the total wind and light output.
Wherein, the pearson correlation coefficient formula is:
wherein ρ is HK Represents the correlation coefficient between the data series H and the data series K, n represents the data number of the data series, H i Represents the ith data, K, in the data series H i Represents the ith data in the data series K,data mean value representing data series H, +.>The data average of data series K is shown.
Specifically, the historical data of the total wind and light output is obtained according to the historical data of the wind power output and the photovoltaic output by taking the day as the data series H in the Pearson correlation coefficient formula, the historical data of the total wind and light output is brought into the data series K in the Pearson correlation coefficient formula, and the correlation coefficient of the load and the total wind and light output is obtained through calculation.
S13, respectively calculating a first correlation coefficient rho between each characteristic quantity data and the peak shaving demand historical data by using a Pearson correlation coefficient formula XY A second correlation coefficient ρ between different feature quantity data XX 。
S2, determining an input variable of a long-term memory neural network prediction model from the characteristic quantity data based on the correlation coefficient among different data, and determining the peak shaving demand historical data as an output variable of the long-term memory neural network prediction model, wherein the method specifically comprises the following steps of:
s21, judging the absolute value |ρ of the first correlation coefficient XY If the I is smaller than or equal to a first preset value, eliminating feature quantity data corresponding to the first correlation coefficient from the feature quantity data to obtain feature quantity data after preliminary screening; if not, determining the characteristic quantity data as the characteristic quantity data after preliminary screening.
In an alternative embodiment, the first preset value is 0.2.
S22, judging the absolute value |ρ of the first correlation coefficient in the feature quantity data after the preliminary screening XY Whether the I is greater than the first preset value and less than the second preset value, if yes, executing S221; and if not, determining the feature quantity data after the preliminary screening as final feature quantity data.
In an alternative embodiment, the second preset value is 0.8.
S221, determining the feature quantity data corresponding to the first correlation coefficient as target feature quantity data, and judging the absolute value |ρ of the second correlation coefficient of the target feature quantity data XX Whether the I is larger than or equal to a third preset value, if so, comparing a first correlation coefficient rho of the target feature quantity data XY Removing the feature quantity data with the minimum absolute value of the first correlation coefficient to obtain final feature quantity data;
in an alternative embodiment, the third preset value is 0.6.
For example, the target feature quantity data is X 1 ，X 2 The absolute value of the second correlation coefficient of (2) is greater than or equal to 0.6, then X is further compared 1 Absolute value of correlation coefficient with peak shaving demand history data Y, X 2 Absolute value of correlation coefficient with peak shaving demand history data Y. Let X be 1 The absolute value of the correlation coefficient with the peak shaving demand historical data Y is minimum, then X is calculated 1 And removing the characteristic quantity data from the preliminarily screened characteristic quantity data.
S23, determining the final characteristic quantity data as an input variable of a long-term and short-term memory neural network prediction model.
S24, determining the peak shaving demand historical data as an output variable of a long-short-period memory neural network prediction model.
S3, training the long-short-term memory neural network prediction model by using the input variable and the output variable to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on the input variable, wherein the method specifically comprises the following steps of:
s31, dividing the input variable and the output variable into four-quarter data according to quarters;
specifically, the time corresponding to the input variable is divided into four-quarter data according to different quarters, and the time corresponding to the output variable is divided into four-quarter data according to different quarters.
S32, randomly extracting one fifth of combinations from each quarter of the four-quarter data of the input variable and each quarter of the four-quarter data of the output variable to form a test set;
for example, a fifth combination is randomly extracted from the first quarter data, the second quarter data, the third quarter data and the fourth quarter data of the four quarter data of the input variable in sequence to form a test set of the input variable, and the test set of the output variable is similar to the above manner.
S33, combining the remaining four fifths in each quarter of the four-quarter data of the input variable and each quarter of the four-quarter data of the output variable to form a training set;
for example, four fifths of a combination are randomly extracted from the first quarter data, the second quarter data, the third quarter data and the fourth quarter data of the four quarter data of the input variable in sequence to form a training set of the input variable, and the training set of the output variable is similar to the above manner.
S34, training the long-short-term memory neural network prediction model by using the test set and the training set to obtain a peak regulation demand prediction model of the power system;
s35, predicting the input variables in the training set or the test set by using the power system peak shaving demand prediction model to obtain a peak shaving demand predicted value.
S4, determining a peak shaving demand real value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data, and performing probability density function fitting on a deviation value between the peak shaving demand predicted value and the peak shaving demand real value to obtain a peak shaving demand predicted interval meeting a confidence level requirement, wherein the peak shaving demand predicted interval specifically comprises the following steps:
s41, determining a peak shaving demand true value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data.
S42, carrying out per unit processing on the peak shaving demand predicted value and the peak shaving demand real value by taking the daily load maximum value as a reference to obtain the processed peak shaving demand predicted value and peak shaving demand real value.
S43, calculating a predicted deviation value between the processed peak shaving demand predicted value and the peak shaving demand true value.
S44, performing fitting calculation on the predicted deviation value by using non-parameter kernel density estimation to obtain a probability density function of the predicted deviation value, and specifically:
wherein f h A probability density function representing the predicted deviation value, N representing the number of data of the predicted deviation value, h representing bandwidth, K (·) representing a kernel function, x representing a deviation value argument, x i Representing the ith sample in the predicted deviation value.
S45, determining the confidence level of the predicted deviation value;
wherein, the confidence level alpha can be set according to actual conditions.
S46, determining a peak shaving demand prediction interval meeting the confidence level requirement based on the peak shaving demand prediction value, the confidence level and the probability density function;
the peak shaving demand prediction interval meeting the confidence level requirement is specifically:
wherein Y is * Representing the peak shaver demand predicted value,indicating a confidence level of +.>When, the inverse function of the distribution function corresponding to the probability density function of the predicted deviation value, +.>Indicating a confidence level of +.>When the inverse function of the distribution function corresponding to the probability density function of the predicted deviation value satisfies +.>
The power system peak shaving demand prediction method is applied to a specific scene:
(1) And acquiring historical data of hour peak regulation demand Y, load, wind power output and photovoltaic output of a provincial power grid within one year.
(2) And extracting characteristic quantity data X affecting peak regulation requirements from historical data of load, wind power output and photovoltaic output by taking a day as a unit to obtain a daily load maximum value X1, a daily load minimum value X2, a wind power output maximum value X3, a wind power output minimum value X4, a photovoltaic output maximum value X5 and a correlation coefficient X6 of total output of load and wind and light.
(3) And calculating the correlation coefficient between each characteristic quantity data in X and peak shaving demand historical data Y and the correlation coefficient between different characteristic quantity data in X, wherein the calculation result is shown in figure 3.
(4) And screening out input variables X1, X5 and X6 of the long-short term memory neural network prediction model based on the correlation coefficient of each characteristic quantity data and the peak shaving demand historical data Y and the correlation coefficient between different characteristic quantity data in the X, wherein the output variable is the peak shaving demand historical data Y.
(5) Training the long-period memory neural network prediction model by using the X1, X5 and X6 and the peak shaving demand historical data Y to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on an input variable (a test set or a training set).
Table 1 compares the influence of the input variable before and after screening on the model prediction accuracy, and it can be seen that after the correlation coefficient screening, the average absolute percentage error, the root mean square error and the residual square sum index of the predicted data can be reduced, and the accuracy of peak shaving demand prediction can be improved.
Table 1 peak shaver demand prediction error coefficients before and after input variable screening
(6) And determining a peak shaving demand true value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data.
(7) And carrying out per unit processing on the peak shaving demand predicted value and the peak shaving demand real value by taking the daily load maximum value as a reference to obtain the processed peak shaving demand predicted value and the peak shaving demand real value.
(8) And calculating a predicted deviation value between the processed peak shaving demand predicted value and the peak shaving demand true value.
(9) And performing fitting calculation on the predicted deviation value by using the non-parameter kernel density estimation to obtain a probability density function of the predicted deviation value.
(10) And determining the confidence level of the predicted deviation value to be 0.8, and determining a peak shaving demand prediction interval meeting the confidence level requirement based on the peak shaving demand predicted value, the confidence level and the probability density function.
As shown in fig. 4, fig. 4 illustrates a comparison between a peak shaving demand prediction value based on a peak shaving demand prediction model of the power system (i.e., a prediction value based on a long-short-term memory neural network in fig. 4) and a peak shaving demand prediction interval (i.e., a true value in fig. 4) meeting a confidence level requirement, where the peak shaving demand prediction interval is substantially within the peak shaving demand prediction value based on the peak shaving demand prediction model of the power system, which illustrates probability distribution characteristics of prediction deviation can be fully considered, so that the adaptability of the prediction result is better, and the accuracy and adaptability of the peak shaving demand prediction are improved.
Example two
Referring to fig. 2, a power system peak shaving demand prediction terminal of the present embodiment includes a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor implements each step in the power system peak shaving demand prediction method of the first embodiment when executing the computer program.
In summary, according to the peak shaving demand prediction method and terminal for the electric power system, correlation analysis is performed based on the characteristic quantity data and the peak shaving demand historical data which affect the peak shaving demand, so that correlation coefficients among different data are obtained, input variables of a model are determined from the characteristic quantity data based on the correlation coefficients among different data, the peak shaving demand historical data are determined to be output variables of the model, the long-short-period memory neural network prediction model is trained by using the characteristic quantity data and the characteristic quantity data, the prediction precision of the model is effectively improved, and finally probability density function fitting is performed on deviation values between peak shaving demand predicted values and peak shaving demand actual values output by the trained prediction model, so that a peak shaving demand prediction interval meeting the confidence level requirement is obtained, probability distribution characteristics of prediction deviation can be fully considered, adaptability of a prediction result is better, and accuracy and adaptability of peak shaving demand prediction are improved; specifically, a pearson correlation coefficient formula is used for respectively calculating a first correlation coefficient between each feature quantity data and the peak shaving demand historical data and a second correlation coefficient between different feature quantity data, the first correlation coefficient can embody the influence degree of each feature quantity data on the peak shaving demand, the second correlation coefficient can embody the mutual influence degree of different feature quantity data, and the first correlation coefficient and the second correlation coefficient are beneficial to the subsequent screening of the feature quantity data most relevant to the peak shaving demand to train a prediction model, so that the accuracy of the prediction model is improved.
The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent changes made by the specification and drawings of the present invention, or direct or indirect application in the relevant art, are included in the scope of the present invention.
Claims (10)
1. The peak regulation demand prediction method for the electric power system is characterized by comprising the following steps of:
acquiring peak shaving demand historical data and characteristic quantity data influencing peak shaving demands of an electric power system, and carrying out correlation analysis based on the characteristic quantity data and the peak shaving demand historical data to obtain correlation coefficients among different data;
determining an input variable of a long-term memory neural network prediction model from the characteristic quantity data based on the correlation coefficient among different data, and determining the peak shaving demand historical data as an output variable of the long-term memory neural network prediction model;
training the long-period and short-period memory neural network prediction model by using the input variable and the output variable to obtain a power system peak shaving demand prediction model, and obtaining a peak shaving demand prediction value by using the power system peak shaving demand prediction model based on the input variable;
and determining a peak shaving demand real value corresponding to the peak shaving demand predicted value from the peak shaving demand historical data, and performing probability density function fitting on a deviation value between the peak shaving demand predicted value and the peak shaving demand real value to obtain a peak shaving demand predicted interval meeting the confidence level requirement.
2. The method for predicting peak shaving demand of a power system according to claim 1, wherein the performing correlation analysis based on the feature quantity data and the peak shaving demand history data to obtain correlation coefficients between different data includes:
and respectively calculating a first correlation coefficient between each characteristic quantity data and the peak shaving demand historical data and a second correlation coefficient between different characteristic quantity data by using a Pearson correlation coefficient formula.
3. The method for predicting peak shaving demand of a power system according to claim 2, wherein determining the input variables of the long-term and short-term memory neural network prediction model from the feature quantity data based on the correlation coefficient between the different data comprises:
judging whether the absolute value of the first correlation coefficient is smaller than or equal to a first preset value, if not, determining the characteristic quantity data as the characteristic quantity data after preliminary screening, and if so, eliminating the characteristic quantity data corresponding to the first correlation coefficient from the characteristic quantity data to obtain the characteristic quantity data after preliminary screening;
judging whether the absolute value of the first correlation coefficient in the preliminarily screened feature quantity data is larger than a first preset value and smaller than a second preset value, if not, determining the preliminarily screened feature quantity data as final feature quantity data, if so, determining the feature quantity data corresponding to the first correlation coefficient as target feature quantity data, judging whether the absolute value of the second correlation coefficient of the target feature quantity data is larger than or equal to a third preset value, if so, comparing the first correlation coefficient of the target feature quantity data, and eliminating the feature quantity data with the minimum absolute value of the first correlation coefficient to obtain final feature quantity data;
and determining the final characteristic quantity data as input variables of a long-term and short-term memory neural network prediction model.
4. The method for predicting peak shaving demand of a power system according to claim 1, wherein training the long-short-term memory neural network prediction model using the input variable and the output variable to obtain the peak shaving demand prediction model of the power system comprises:
dividing the input variable and the output variable into four-quarter data according to quarters respectively;
randomly extracting one fifth combination from each quarter of the four-quarter data of the input variable and each quarter of the four-quarter data of the output variable to form a test set;
combining the remaining four fifths of the four quarters of the input variable and the four quarters of the output variable to form a training set;
training the long-short-period memory neural network prediction model by using the test set and the training set to obtain a peak regulation demand prediction model of the power system;
the obtaining a peak shaving demand prediction value based on the input variable using the power system peak shaving demand prediction model includes:
and predicting the input variables in the training set or the test set by using the power system peak shaving demand prediction model to obtain a peak shaving demand prediction value.
5. The power system peak shaver demand prediction method according to claim 1, wherein the characteristic amount data affecting the peak shaver demand includes a daily load maximum value;
and performing probability density function fitting on the deviation value between the peak shaving demand predicted value and the peak shaving demand true value to obtain a peak shaving demand predicted interval meeting the confidence level requirement, wherein the peak shaving demand predicted interval comprises the following steps:
performing per unit processing on the peak shaving demand predicted value and the peak shaving demand real value by taking the daily load maximum value as a reference to obtain the processed peak shaving demand predicted value and peak shaving demand real value;
calculating a predicted deviation value between the processed peak shaving demand predicted value and the peak shaving demand true value;
performing fitting calculation on the predicted deviation value by using non-parameter kernel density estimation to obtain a probability density function of the predicted deviation value;
determining a confidence level of the predicted deviation value;
and determining a peak shaving demand prediction interval meeting the confidence level requirement based on the peak shaving demand prediction value, the confidence level and the probability density function.
6. The method of claim 5, wherein the performing a fitting calculation on the predicted deviation value using a non-parametric kernel density estimate to obtain a probability density function of the predicted deviation value comprises:
wherein f h A probability density function representing the predicted deviation value, N representing the number of data of the predicted deviation value, h representing bandwidth, K (·) representing a kernel function, x representing a deviation value argument, x i Representing the ith sample in the predicted deviation value.
7. The method for predicting peak shaver demand of a power system according to claim 1, wherein the obtaining the peak shaver demand history data and the characteristic amount data affecting the peak shaver demand of the power system comprises:
acquiring peak shaving demand historical data of an electric power system and historical data which corresponds to the peak shaving demand historical data and affects peak shaving demands;
and extracting characteristic quantity data affecting the peak shaving demand from the historical data affecting the peak shaving demand by taking the day as a unit.
8. The method for predicting peak shaver demand in an electric power system according to claim 7, wherein the historical data affecting the peak shaver demand includes historical data of load, wind power output and photovoltaic output;
the characteristic quantity data influencing peak shaving demands comprise a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value, a photovoltaic output maximum value and a correlation coefficient of load and wind and light total output;
the extracting, from the historical data affecting the peak shaving demand in units of days, feature quantity data affecting the peak shaving demand includes:
respectively counting historical data of the load, the wind power output and the photovoltaic output by taking a day as a unit to obtain a daily load maximum value, a daily load minimum value, a wind power output maximum value, a wind power output minimum value and a photovoltaic output maximum value;
and calculating according to historical data of the load, wind power output and photovoltaic output by using a pearson correlation coefficient formula by taking a day as a unit to obtain a correlation coefficient of the load and the total wind and light output.
9. The power system peak shaving demand prediction method according to claim 2 or 8, wherein the pearson correlation coefficient formula is:
wherein ρ is HK Represents the correlation coefficient between the data series H and the data series K, n represents the data number of the data series, H i Represents the ith data, K, in the data series H i Represents the ith data in the data series K,data mean value representing data series H, +.>The data average of data series K is shown.
10. A power system peak shaving demand prediction terminal comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of a power system peak shaving demand prediction method according to any one of claims 1 to 9 when the computer program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310674840.5A CN116701869A (en) | 2023-06-08 | 2023-06-08 | Peak regulation demand prediction method and terminal for power system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310674840.5A CN116701869A (en) | 2023-06-08 | 2023-06-08 | Peak regulation demand prediction method and terminal for power system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116701869A true CN116701869A (en) | 2023-09-05 |
Family
ID=87828745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310674840.5A Pending CN116701869A (en) | 2023-06-08 | 2023-06-08 | Peak regulation demand prediction method and terminal for power system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116701869A (en) |
-
2023
- 2023-06-08 CN CN202310674840.5A patent/CN116701869A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111027775A (en) | Step hydropower station generating capacity prediction method based on long-term and short-term memory network | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN103971175B (en) | Short-term load prediction method of multistage substations | |
| CN112308427A (en) | New energy consumption restriction factor evaluation method and system based on combined empowerment-grey correlation | |
| CN110212592B (en) | Thermal power generating unit load regulation maximum rate estimation method and system based on piecewise linear expression | |
| CN117744916A (en) | Method and device for predicting energy storage capacity, computer equipment and readable storage medium | |
| CN115640874A (en) | Transformer state prediction method based on improved grey model theory | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
| CN113919594A (en) | Demand response potential evaluation method based on deep forest | |
| CN114037209A (en) | Comprehensive benefit analysis method and device for distributed photovoltaic access direct-current power distribution system | |
| CN110210755B (en) | User demand response capability assessment method based on K _ means clustering algorithm | |
| CN116701869A (en) | Peak regulation demand prediction method and terminal for power system | |
| CN116470491A (en) | Photovoltaic power probability prediction method and system based on copula function | |
| CN114266423B (en) | Multi-model-based power energy demand prediction method and device and storage medium | |
| CN115967092A (en) | Data-driven non-parameter probability optimal power flow prediction-analysis integrated method for new energy power system | |
| Wilson et al. | Use of meteorological data for improved estimation of risk in capacity adequacy studies | |
| CN112668770B (en) | Power load prediction method based on overall similarity of information and waveforms | |
| CN110322063B (en) | Power consumption simulation prediction method and storage medium | |
| Yanhui et al. | Forecasting hourly electricity demand with nonparametric functional data analysis | |
| CN117526316B (en) | Load prediction method based on GCN-CBAM-BiGRU combined model | |
| US20240195174A1 (en) | Energy prediction device, energy prediction method, and energy prediction program | |
| Li et al. | Application of improved gm (1, n) models in annual electricity demand forecasting | |
| CN116128547A (en) | Method and system for predicting current price of electric power spot market based on deep AR model | |
| Xu et al. | Short-term electricity load forecasting based on ensemble empirical mode decomposition and long short-term memory neural network | |
| CN117875467A (en) | Power system payload prediction method, device, electronic equipment and storage medium |
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 |