CN105701570A - Short-term electric power demand analysis method based on overall process technology improvement - Google Patents

Short-term electric power demand analysis method based on overall process technology improvement Download PDF

Info

Publication number
CN105701570A
CN105701570A CN201610016411.9A CN201610016411A CN105701570A CN 105701570 A CN105701570 A CN 105701570A CN 201610016411 A CN201610016411 A CN 201610016411A CN 105701570 A CN105701570 A CN 105701570A
Authority
CN
China
Prior art keywords
load
analysis
day
meteorological
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201610016411.9A
Other languages
Chinese (zh)
Inventor
吴臻
兰洲
邢胜男
戴攀
沈志恒
石清
李黎
张婕
孙飞飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BEIJING JINGSHI WANFANG INFORMATION TECHNOLOGY Co Ltd
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
BEIJING JINGSHI WANFANG INFORMATION TECHNOLOGY Co Ltd
State Grid Corp of China SGCC
Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by BEIJING JINGSHI WANFANG INFORMATION TECHNOLOGY Co Ltd, State Grid Corp of China SGCC, Economic and Technological Research Institute of State Grid Zhejiang Electric Power Co Ltd filed Critical BEIJING JINGSHI WANFANG INFORMATION TECHNOLOGY Co Ltd
Priority to CN201610016411.9A priority Critical patent/CN105701570A/en
Publication of CN105701570A publication Critical patent/CN105701570A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06QDATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation, e.g. linear programming, "travelling salesman problem" or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06QDATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06Electricity, gas or water supply

Abstract

The present invention discloses a short-term electric power demand analysis method based on overall process technology improvement. The short-term load predication is an important component of the load prediction, and has important meaning for schedulling and arranging turn-on and turn-off plans and for the electric power application, such as the unit optimal combination, the economic dispatch, the optimal power flow, etc. The short-term electric power demand analysis method of the present invention enables the short-term electric power demand before/at/after prediction overall process technologies to be integrated and improved, a before-prediction technology comprises the data pre-processing and the data analysis, and a data pre-processing method comprises intelligently identifying and correcting the bad data, removing the influences of holidays and festivals, removing the load natural increase influence and fully considering the accumulative effect of the recent history data. An at-prediction technology comprises meteorological influence effect analysis and a load prediction model set, and an after-prediction technology comprises a prediction evaluation mechanism. The method of the present invention helps perfect a short-term electric power demand prediction technology and enables the short-term load prediction precision to be improved.

Description

A kind of short term power requirement analysis method based on overall process skill upgrading
Technical field
The present invention relates to electricity needs analysis field, especially a kind of short term power requirement analysis method based on overall process skill upgrading。
Background technology
Short-term load forecasting is the important component part of load prediction; it is mainly used in the electric load in forecast coming few minutes, several hours or a few week; for schedule startup-shutdown plan, electric power application such as unit commitment, economic load dispatching, optimal load flow are had great significance。Consumption habit in load response people productive life, and the activity of people can be subject to the impact of day and night change, seasonal law quasi-periodic, development especially as society has found out the habit meeting people gradually to improve production efficiency, such as working day and day off etc., which results in the productive life of people and have stronger regularity。The factors such as the sudden change of weather simultaneously, festivals or holidays or government policy also can affect workload demand and produce change。Produce and regularity in one's life it follows that the change of load depends primarily on people, and be subject to the impact of some correlative factors (such as temperature, rain or shine sleet etc.)。Random factor in load variations is objective reality again, any exquisite load forecasting method can not be completely eliminated error, therefore the task of load prediction is exactly that the forecast model adopting science fully excavates the regularity in demand history data, consider the comprehensive of influence factor's information, thus reducing prediction error as much as possible。
Summary of the invention
It is an object of the invention to provide a kind of short term power requirement analysis method based on overall process skill upgrading, contribute to improving short term power requirement forecasting technology, improve short-term load forecasting precision。
For this, the present invention adopts the following technical scheme that: it is characterized in that, overall process technology before, during and after short term power requirement forecasting is integrated and is promoted by it, before described prediction, technology includes data prediction and data analysis, data prediction includes bad data INTELLIGENT IDENTIFICATION and correction, the removal of impact festivals or holidays, the removal of load natural increase impact and takes into full account the accumulative effect of historical data recently, data analysis include day, week, the moon, season, year load Analysis, stability analysis and set up typical curve storehouse;In described prediction, technology includes meteorological effect effect analysis and load forecasting model set, and after described prediction, technology includes forecast assessment mechanism。
Further, the particular content of described data prediction is as follows:
1) bad data INTELLIGENT IDENTIFICATION and correction: first slightly identify, to exceptional data point simple modifications, then characteristic curve is extracted, relative analysis characteristic curve and load curve, secondly it is modified according to characteristic curve, realize fine identification abnormal data, finally revised data are carried out secondary check, the data of erroneous judgement are reduced;
2) analysis and research time to distinguish working day and festivals or holidays load;
3) formative factor of daily load is considered, daily load is divided into three parts constitute, is load natural increase part, meteorological load and random load respectively, eliminates the impact of load natural increase factor, ignore the impact of random load, extract meteorological load as analyzing object;
4) according to the principle of " near big and far smaller " to meteorological factor synthesization in many days, the accumulative effect of first some days is taken into full account。
Further, the particular content of described data analysis is as follows:
1) daily load analysis includes enumerating the analysis of type load curve, the analysis of sustained load curve, load curve contrast, daily load specificity analysis, the analysis of certain period load trend, certain period part throttle characteristics trend analysis and the relative analysis of many daily loads;
2) week load Analysis includes week load curve analysis, all Load Characteristic Analysis, all part throttle characteristics trend analysiss, week exemplary operation day tracing analysis and week typical case tracing analysis on day off;
3) moon load Analysis include a moon Load Characteristic Analysis, the moon part throttle characteristics trend analysis, the exemplary operation moon, tracing analysis day, the tracing analysis Saturday moon, tracing analysis on the Sunday moon, the moon maximum electricity day tracing analysis, the minimum amount of power moon, tracing analysis day, the tracing analysis of maximum monthly load day, the minimum load moon, tracing analysis day, the moon maximum peak-valley difference day tracing analysis and the moon minimum peak-valley difference day tracing analysis;
4) season load Analysis include season Load Characteristic Analysis, season part throttle characteristics trend analysis, exemplary operation season, tracing analysis day, tracing analysis Saturday in season, tracing analysis on Sunday in season, season maximum electricity day tracing analysis, minimum amount of power season, tracing analysis day, season peak day tracing analysis, minimum load season, tracing analysis day, season maximum peak-valley difference day tracing analysis and season minimum peak-valley difference day tracing analysis;
5) year load Analysis include yearly load curve analysis, year Load Characteristic Analysis, year part throttle characteristics trend analysis, year lasting load curve analysis and load probabilistic distribution analysis;
6) stability analysis is to be undertaken decomposing and component analysis by the data of aim curve a period of time, extracting wherein can the composition of regular modeling and undulatory property composition, so that it is determined that the regularity of this section of historical data and predictability, the possible precision of assessment prediction, method choice and precision of prediction Pre-Evaluation for subsequent prediction flow process provide reference;
7) arranging season and monthly typical curve in typical curve storehouse, it is named operation and preserves, predicting the outcome, correction link recalls set typical curve, until revising reference role。
Further, in described prediction, the particular content of technology is as follows:
1) meteorological effect effect analysis: network load and single meteorological index dependency carry out numerical analysis, and research is applicable to the comprehensive meteorological index constructive method of network load prediction。On this basis, research different times affects the leading meteorological factor of load the influence degree that this meteorological index of quantitative analysis is to part throttle characteristics。
11) correlation analysis and leading meteorological factor identification
The method taking quantitative analysis, describes the dependency between meteorological index and load using the correlation coefficient between load and meteorological factor as quantizating index, so that it is determined that leading meteorological factor。Load index include Daily treatment cost, per day load, day minimum load and day electricity etc., meteorological index divides single meteorological index and comprehensive meteorological index, the former includes mean temperature, maximum temperature, minimum temperature, rainfall, relative humidity and wind speed index, and the latter includes sendible temperature and human comfort index。
12) comprehensive meteorological index is based on ride number, and it is the meteorological element comprehensive functions to human body such as tolerance temperature, humidity, wind speed, characterizes human body whether comfortable in atmospheric environment, and its computing formula is:
k = 1.8 T a - 0.55 ( 1 - R h ) - 3.2 V + 32
In formula, k is Body Comfort Index;TaIt is temperature (DEG C);RhIt is relative humidity (%);V is wind speed (m/s)。
Following table is the grade drawn according to Body Comfort Index and human body is felt accordingly。
Table 1 human comfort table corresponding to human body sensory
k Index ranking Human body sensory describes
<25 4 grades Human body sensory is cold, extremely inadaptable
26~38 3 grades Human body sensory is terribly cold, very uncomfortable
39~50 2 grades Human body sensory is colder, uncomfortable
51~58 1 grade Human body sensory is slightly biased cool, comparatively comfortable
59~70 0 grade Human body sensory is the most comfortable, the most acceptable
71~75 1 grade Human body sensory is partially warm, comparatively comfortable
76~79 2 grades Human body sensory partial heat, uncomfortable
80~85 3 grades Human body sensory is hot, very uncomfortable
86~89 4 grades Human body sensory is awfully hot, extremely uncomfortable
>90 5 grades Human body sensory is extremely hot, needs preventing heatstroke, in case heatstroke
Wherein wind speed can obtain according to wind scale, it may be assumed that
V=0.836 × B1.5
In formula, V is wind speed, and B is wind scale。
13) meteorological sensitive analysis
Meteorological sensitive analysis refers to the sensitivity that certain meteorological index is changed by load, namely in the rate of change size to this meteorological index of the lower load with certain meteorological index best fit。It is predicted as example with peak load and maximum temperature variable quantity, needs during sensitive analysis first to collect the Daily treatment cost in certain year certain season and maximum temperature data, describe its scatterplot;Secondly scatterplot being carried out curve fitting, the mode of matching includes, up to 11 kinds, finding out its optimal matched curve, such as 3 regression curves:
P (x)=a0+a1x1+a2x2+…+anxn
P (x) is peak load index, and x is maximum temperature index。DP (x)/dx is the Daily treatment cost sensitive rate of change with maximum temperature, the i.e. load sensitivity to Meteorological Index, it is also possible to think the variable quantity at the Daily treatment cost caused by unit maximum temperature index variation。Here take piecewise approximation and describe the mode of curve, it is possible to the Daily treatment cost variable quantity under certain maximum temperature is made reliable estimation。
Consider the difference of province's electrical network each region meteorological factor and load, it is possible to calculate its meteorological factor quantification index respectively by each department meteorological data, draw the quantification index value of province's electrical network further according to the load meteorology sensitivity size weighting of each department, namely
T=∑ Ti(ΔEmi/∑ΔEmi)
In formula, Ti is each department Meteorological Index, and Δ Emi is each department sensitivity numerical value。
So that the result of sensitive analysis more science is reliable, still uses multi-target analysis and select and the method on reserve section day off holiday is analyzed。The process analyzed also can add Thermal incubation effect so that analyze result more accurately reliable。
2) load forecasting model set:
21) based on the conventional forecast model of same type day, selection and the day to be predicted historical load day with type, analyze its Changing Pattern, arrange near to the order farthest away from day to be predicted, extract the Daily treatment cost with type day, day minimum load, per day load, and in this, as reference value, to this daily load curve standardization, adopt point-to-point multiple proportions method, equal lotus mode multiple proportions smoothing techniques, paddy lotus mode multiple proportions smoothing techniques, peak load mode multiple proportions smoothing techniques, equal lotus mode overlap smoothing techniques, paddy lotus mode overlap smoothing techniques, peak load mode overlap smoothing techniques, Linear regression is analyzed prediction。
22) newly cease forecast model, adopt new breath (namely up-to-date information) that following daily load is predicted。Utilize new breath to utilize incomplete information prediction on the same day time daily load, rather than adopt complete information on load the previous day。By the same day known information on load as datum mark, one virtual sky of 24 hours of manual construction, thus realizing the utilization of latest data, reach the purpose of prediction。To based on same type day model many algorithms improve, form new breath forecast model class algorithm: new cease point-to-point multiple proportions method, new cease equal lotus mode multiple proportions smoothing techniques, new breath paddy lotus mode multiple proportions smoothing techniques, new breath peak load mode multiple proportions smoothing techniques, new cease equal lotus mode overlap smoothing techniques, new breath paddy lotus mode overlap smoothing techniques, new breath peak load mode overlap smoothing techniques。
23) based on the forecast model of frequency domain decomposition method: electrically-based load has relatively strong periodically, it is used for time series frequency-domain analysis method to be predicted, following Fourier decomposition can be done at Load Time Series P (t) specifying modeling time domain D-:
P ( t ) = a 0 + &Sigma; i = 1 N - 1 ( a i cos&omega; i t + b i sin&omega; i t )
Wherein, N is the length of load sequence。
Character according to Fourier decomposition, the signal obtained after decomposition is mutually orthogonal directions。Load P (t) can resolve into angular frequency in this way isComponent。By suitable combination, and according to the periodic feature of load variations, P (t) can be reconstructed following formula:
P (t)=a0+D(t)+W(t)+L(t)+H(t)
Wherein, diurnal periodicity component a0+ D (t) and week periodic component W (t) are the load components by fixed cycle change。The algorithm of this model has low frequency average frequency domain component method, low frequency to smooth frequency domain components method, the relevant frequency domain components method of low frequency, the relevant frequency domain components method of surplus。
24) forecast model of meteorological factor is considered: the consideration meteorological factor forecast model based on mapping library set up for solving different factor dimension difference。On the one hand the original quantitative targets such as temperature, rainfall, wind speed, relative humidity are mapped, on the other hand day weather category (fine, rain, cloudy etc.), week type (Monday, week second-class), date poor (history day with predict day number of days different: 1 day, 2 days etc.) are mapped。The setting of different mappings value can be carried out according to the actual weather in grid company location。Pattern-recongnition method, similarity extrapolation, correlative factor matching method is adopted to carry out load prediction。Artificial neural network method is may be used without not carrying out when mapping library is arranged。
25) special forecast model festivals or holidays: the model built for country's legal major holiday or holiday, generally comprises festivals or holidays such as the Spring Festival, New Year's Day, Clear and Bright, the Dragon Boat Festival, International Labour Day, National Day etc.。Although festivals or holidays, load curve differed relatively big with working day, but the load curve of longitudinal comparison same festivals or holidays is it can be seen that it is quite similar。Thus select and based on the conventional forecast model thought of same type day, it be predicted, concrete grammar have equal lotus mode multiple proportions smoothing techniques festivals or holidays, festivals or holidays paddy lotus mode multiple proportions smoothing techniques, festivals or holidays peak load mode multiple proportions smoothing techniques, pointwise festivals or holidays growth rate method。
Further, described forecast assessment mechanism refers to carry out load index predictive value and actual value proof analysis, the practical effect of checking model, adds up the whole network load prediction accuracy rate in many daily ranges, accuracy rate ranking and the contrast same period。
The present invention quantizes identification and evaluation methodology based on the load regularity of degree of stability, effectively promote network load precision of prediction level, promote electrical network lean, standardized management level, promote power grid security, high-quality and economical operation, provide technology guarantee for carrying out energy-saving power generation dispatching work。
Accompanying drawing explanation
Fig. 1 is principles of the invention figure。
Detailed description of the invention
The present invention integrates the overall process technology before, during and after short term power requirement forecasting, contains the technology such as data prediction, data analysis, prediction algorithm, predicting strategy and forecast assessment mechanism。Before described prediction, technology includes data prediction and data analysis。
1, in the present invention, it was predicted that front data prediction includes bad data INTELLIGENT IDENTIFICATION and correction, the removal of impact festivals or holidays, the removal of load natural increase impact and takes into full account the accumulative effect of historical data recently。
11) bad data INTELLIGENT IDENTIFICATION and modification method are first slightly to identify, to exceptional data point simple modifications, then characteristic curve is extracted, relative analysis characteristic curve and load curve, secondly it is modified according to characteristic curve, realize fine identification abnormal data, finally revised data are carried out secondary check, the data of erroneous judgement are reduced。
12) due to festivals or holidays load with obvious difference on working day, and relatively small, to distinguish during analysis and research working day and festivals or holidays load。
13) consider the formative factor of daily load, it is believed that daily load is made up of three parts, be load natural increase part, meteorological load and random load respectively。System should be able to eliminate the impact of load natural increase factor, ignores the impact of random load, extracts meteorological load as analyzing object。
14) according to the principle of " near big and far smaller " to meteorological factor synthesization in many days, the accumulative effect of first some days is taken into full account。
2, in the present invention, it was predicted that front data analysis include day, week, the moon, season, year load Analysis, stability analysis and set up typical curve storehouse。
21) daily load analysis includes enumerating the analysis of type load curve, the analysis of sustained load curve, load curve contrast, daily load specificity analysis, the analysis of certain period load trend, certain period part throttle characteristics trend analysis, the relative analysis of many daily loads。
22) week load Analysis includes week load curve analysis, all Load Characteristic Analysis, all part throttle characteristics trend analysiss, week exemplary operation day tracing analysis, week typical case's tracing analysis on day off。
23) moon (season) load Analysis includes the moon (season) Load Characteristic Analysis, month (season) part throttle characteristics trend analysis, the moon (season) exemplary operation day tracing analysis, the moon (season) tracing analysis Saturday, the moon (season) tracing analysis on Sunday, the moon (season) maximum electricity day tracing analysis, the moon (season) minimum amount of power day tracing analysis, the moon (season) peak day tracing analysis, the moon (season) minimum load day tracing analysis, the moon (season) maximum peak-valley difference day tracing analysis, the moon (season) minimum peak-valley difference day tracing analysis。
24) year load Analysis include yearly load curve analysis, year Load Characteristic Analysis, year part throttle characteristics trend analysis, year lasting load curve analyze, load probabilistic distribution analysis。
25) stability analysis is to be undertaken decomposing and component analysis by the data of aim curve a period of time, extracting wherein can the composition of regular modeling and undulatory property composition, so that it is determined that the regularity of this section of historical data and predictability, the possible precision of assessment prediction, method choice and precision of prediction Pre-Evaluation for subsequent prediction flow process provide reference。
26) typical curve storehouse arranges season and monthly typical curve, it is named operation and preserves。May bring up set typical curve in the correction link that predicts the outcome and be modified reference。
3, in the present invention, it was predicted that middle technology includes meteorological effect effect analysis, load forecasting model set。
31) meteorological effect effect analysis: network load and single meteorological index dependency carry out numerical analysis, and research is applicable to the comprehensive meteorological index constructive method of network load prediction。On this basis, research different times affects the leading meteorological factor of load the influence degree that this meteorological index of quantitative analysis is to part throttle characteristics。
311) correlation analysis and leading meteorological factor identification
The method taking quantitative analysis, describes the dependency between meteorological index and load using the correlation coefficient between load and meteorological factor as quantizating index, so that it is determined that leading meteorological factor。Load index include Daily treatment cost, per day load, day minimum load and day electricity etc., meteorological index includes the single meteorological index such as mean temperature, maximum temperature, minimum temperature, rainfall, relative humidity and wind speed, also includes the comprehensive meteorological index such as sendible temperature, human comfort。
312) comprehensive meteorological index is based on ride number, and it is the meteorological element comprehensive functions to human body such as tolerance temperature, humidity, wind speed, characterizes human body whether comfortable in atmospheric environment, and its computing formula is:
k = 1.8 T a - 0.55 ( 1 - R h ) - 3.2 V + 32
In formula, k is Body Comfort Index;TaIt is temperature (DEG C);RhIt is relative humidity (%);V is wind speed (m/s)。
Following table is the grade drawn according to Body Comfort Index and human body is felt accordingly。
Table 1 human comfort table corresponding to human body sensory
k Index ranking Human body sensory describes
<25 4 grades Human body sensory is cold, extremely inadaptable
26~38 3 grades Human body sensory is terribly cold, very uncomfortable
39~50 2 grades Human body sensory is colder, uncomfortable
51~58 1 grade Human body sensory is slightly biased cool, comparatively comfortable
59~70 0 grade Human body sensory is the most comfortable, the most acceptable
71~75 1 grade Human body sensory is partially warm, comparatively comfortable
76~79 2 grades Human body sensory partial heat, uncomfortable
80~85 3 grades Human body sensory is hot, very uncomfortable
86~89 4 grades Human body sensory is awfully hot, extremely uncomfortable
>90 5 grades Human body sensory is extremely hot, needs preventing heatstroke, in case heatstroke
Wherein, wind speed can obtain according to wind scale, it may be assumed that
V=0.836 × B1.5
In formula, V is wind speed, and B is wind scale。
313) meteorological sensitive analysis
Meteorological sensitive analysis refers to the sensitivity that certain meteorological index is changed by load, namely in the rate of change size to this meteorological index of the lower load with certain meteorological index best fit。It is predicted as example with peak load and maximum temperature variable quantity, needs during sensitive analysis first to collect the Daily treatment cost in certain year certain season and maximum temperature data, describe its scatterplot;Secondly scatterplot being carried out curve fitting, the mode of matching includes, up to 11 kinds, finding out its optimal matched curve, such as 3 regression curves:
P (x)=a0+a1x1+a2x2+…+anxn
P (x) is peak load index, and x is maximum temperature index。DP (x)/dx is the Daily treatment cost sensitive rate of change with maximum temperature, the i.e. load sensitivity to Meteorological Index, it is also possible to think the variable quantity at the Daily treatment cost caused by unit maximum temperature index variation。Here take piecewise approximation and describe the mode of curve, it is possible to the Daily treatment cost variable quantity under certain maximum temperature is made reliable estimation。
Consider the difference of province's electrical network each region meteorological factor and load, it is possible to calculate its meteorological factor quantification index respectively by each department meteorological data, draw the quantification index value of province's electrical network electrical network further according to the load meteorology sensitivity size weighting of each department, namely
T=∑ Ti(ΔEmi/∑ΔEmi)
Formula ZhongTiWei Ge district Meteorological Index, ΔEmiWei Ge district sensitivity numerical value。
So that the result of sensitive analysis more science is reliable, still uses multi-target analysis and holiday no collection day to select and the method for reservation is analyzed。The process analyzed also can add Thermal incubation effect so that analyze result more accurately reliable。
32) load forecasting model set:
321) based on the conventional forecast model of same type day, selection and the day to be predicted historical load day with type, analyze its Changing Pattern, arrange near to the order farthest away from day to be predicted, extract the Daily treatment cost with type day, day minimum load, per day load, and in this, as reference value, to this daily load curve standardization, adopt point-to-point multiple proportions method, equal lotus mode multiple proportions smoothing techniques, paddy lotus mode multiple proportions smoothing techniques, peak load mode multiple proportions smoothing techniques, equal lotus mode overlap smoothing techniques, paddy lotus mode overlap smoothing techniques, peak load mode overlap smoothing techniques, Linear regression is analyzed prediction。
322) newly cease forecast model, adopt new breath (namely up-to-date information) that following daily load is predicted。Utilize new breath to utilize incomplete information prediction on the same day time daily load, rather than adopt complete information on load the previous day。By the same day known information on load as datum mark, one virtual sky of 24 hours of manual construction, thus realizing the utilization of latest data, reach the purpose of prediction。To based on same type day model many algorithms improve, form new breath forecast model class algorithm: new cease point-to-point multiple proportions method, new cease equal lotus mode multiple proportions smoothing techniques, new breath paddy lotus mode multiple proportions smoothing techniques, new breath peak load mode multiple proportions smoothing techniques, new cease equal lotus mode overlap smoothing techniques, new breath paddy lotus mode overlap smoothing techniques, new breath peak load mode overlap smoothing techniques。
323) based on the forecast model of frequency domain decomposition method: electrically-based load has relatively strong periodicity, is used for time series frequency-domain analysis method and is predicted, modeling time domain D is being specified-Load Time Series P (t) following Fourier decomposition can be done:
P ( t ) = a 0 + &Sigma; i = 1 N - 1 ( a i cos&omega; i t + b i sin&omega; i t )
Wherein, N is the length of load sequence。
Character according to Fourier decomposition, the signal obtained after decomposition is mutually orthogonal directions。Load P (t) can resolve into angular frequency in this way isComponent。By suitable combination, and according to the periodic feature of load variations, P (t) can be reconstructed following formula:
P (t)=a0+D(t)+W(t)+L(t)+H(t)
Wherein, diurnal periodicity component a0+ D (t) and week periodic component W (t) are the load components by fixed cycle change。The algorithm of this model has low frequency average frequency domain component method, low frequency to smooth frequency domain components method, the relevant frequency domain components method of low frequency, the relevant frequency domain components method of surplus。
324) forecast model of meteorological factor is considered: the consideration meteorological factor forecast model based on mapping library set up for solving different factor dimension difference。On the one hand the original quantitative targets such as temperature, rainfall, wind speed, relative humidity are mapped, on the other hand day weather category (fine, rain, cloudy etc.), week type (Monday, week second-class), date poor (history day with predict day number of days different: 1 day, 2 days etc.) are mapped。The setting of different mappings value can be carried out according to the actual weather in grid company location。Pattern-recongnition method, similarity extrapolation, correlative factor matching method is adopted to carry out load prediction。Artificial neural network method is may be used without not carrying out when mapping library is arranged。
325) special forecast model festivals or holidays: the model built for country's legal major holiday or holiday, generally comprises festivals or holidays such as the Spring Festival, New Year's Day, Clear and Bright, the Dragon Boat Festival, International Labour Day, National Day etc.。Although festivals or holidays, load curve differed relatively big with working day, but the load curve of longitudinal comparison same festivals or holidays is it can be seen that it is quite similar。Thus select and based on the conventional forecast model thought of same type day, it be predicted, concrete grammar have equal lotus mode multiple proportions smoothing techniques festivals or holidays, festivals or holidays paddy lotus mode multiple proportions smoothing techniques, festivals or holidays peak load mode multiple proportions smoothing techniques, pointwise festivals or holidays growth rate method。
4, in the present invention, it was predicted that after evaluation mechanism refer to carry out load index predictive value and actual value proof analysis, the practical effect of checking model, add up the whole network load prediction accuracy rate in many daily ranges, accuracy rate ranking and the contrast same period。

Claims (5)

1. the short term power requirement analysis method based on overall process skill upgrading, it is characterized in that, overall process technology before, during and after short term power requirement forecasting is integrated and is promoted by it, before described prediction, technology includes data prediction and data analysis, data prediction includes bad data INTELLIGENT IDENTIFICATION and correction, the removal of impact festivals or holidays, the removal of load natural increase impact and takes into full account the accumulative effect of historical data recently, data analysis include day, week, the moon, season, year load Analysis, stability analysis and set up typical curve storehouse;In described prediction, technology includes meteorological effect effect analysis and load forecasting model set, and after described prediction, technology includes forecast assessment mechanism。
2. short term power requirement analysis method according to claim 1, it is characterised in that the particular content of described data prediction is as follows:
1) bad data INTELLIGENT IDENTIFICATION and correction: first slightly identify, to exceptional data point simple modifications, then characteristic curve is extracted, relative analysis characteristic curve and load curve, secondly it is modified according to characteristic curve, realize fine identification abnormal data, finally revised data are carried out secondary check, the data of erroneous judgement are reduced;
2) analysis and research time to distinguish working day and festivals or holidays load;
3) formative factor of daily load is considered, daily load is divided into three parts constitute, is load natural increase part, meteorological load and random load respectively, eliminates the impact of load natural increase factor, ignore the impact of random load, extract meteorological load as analyzing object;
4) according to the principle of " near big and far smaller " to meteorological factor synthesization in many days, the accumulative effect of first some days is taken into full account。
3. short term power requirement analysis method according to claim 1, it is characterised in that the particular content of described data analysis is as follows:
1) daily load analysis includes enumerating the analysis of type load curve, the analysis of sustained load curve, load curve contrast, daily load specificity analysis, the analysis of certain period load trend, certain period part throttle characteristics trend analysis and the relative analysis of many daily loads;
2) week load Analysis includes week load curve analysis, all Load Characteristic Analysis, all part throttle characteristics trend analysiss, week exemplary operation day tracing analysis and week typical case tracing analysis on day off;
3) moon load Analysis include a moon Load Characteristic Analysis, the moon part throttle characteristics trend analysis, the exemplary operation moon, tracing analysis day, the tracing analysis Saturday moon, tracing analysis on the Sunday moon, the moon maximum electricity day tracing analysis, the minimum amount of power moon, tracing analysis day, the tracing analysis of maximum monthly load day, the minimum load moon, tracing analysis day, the moon maximum peak-valley difference day tracing analysis and the moon minimum peak-valley difference day tracing analysis;
4) season load Analysis include season Load Characteristic Analysis, season part throttle characteristics trend analysis, exemplary operation season, tracing analysis day, tracing analysis Saturday in season, tracing analysis on Sunday in season, season maximum electricity day tracing analysis, minimum amount of power season, tracing analysis day, season peak day tracing analysis, minimum load season, tracing analysis day, season maximum peak-valley difference day tracing analysis and season minimum peak-valley difference day tracing analysis;
5) year load Analysis include yearly load curve analysis, year Load Characteristic Analysis, year part throttle characteristics trend analysis, year lasting load curve analysis and load probabilistic distribution analysis;
6) stability analysis is to be undertaken decomposing and component analysis by the data of aim curve a period of time, and extracting wherein can the regular composition modeled and undulatory property composition;
7) arranging season and monthly typical curve in typical curve storehouse, it is named operation and preserves, predicting the outcome, correction link recalls set typical curve。
4. short term power requirement analysis method according to claim 1, it is characterised in that in described prediction, the particular content of technology is as follows:
1) meteorological effect effect analysis: network load and single meteorological index dependency are carried out numerical analysis, research is applicable to the comprehensive meteorological index constructive method of network load prediction, on this basis, research different times affects the leading meteorological factor of load the influence degree that this meteorological index of quantitative analysis is to part throttle characteristics;
11) correlation analysis and leading meteorological factor identification
The method taking quantitative analysis, describes the dependency between meteorological index and load using the correlation coefficient between load and meteorological factor as quantizating index, so that it is determined that leading meteorological factor;Load index include Daily treatment cost, per day load, day minimum load and day electricity, meteorological index divides single meteorological index and comprehensive meteorological index, the former includes mean temperature, maximum temperature, minimum temperature, rainfall, relative humidity and wind speed index, and the latter includes sendible temperature and human comfort index;
12) comprehensive meteorological index is based on Body Comfort Index, and it is tolerance temperature, humidity, the comprehensive function to human body of the wind speed meteorological element, characterizes human body whether comfortable in atmospheric environment, and its computing formula is:
k = 1.8 T a - 0.55 ( 1 - R h ) - 3.2 V + 32 ,
In formula, k is Body Comfort Index;TaIt is temperature, DEG C;RhIt is relative humidity, %;V is wind speed, m/s;
Wherein, wind speed obtains according to wind scale, it may be assumed that
V=0.836 × B1.5,
In formula, V is wind speed, and B is wind scale;
13) meteorological sensitive analysis
Meteorological sensitive analysis refers to the sensitivity that certain meteorological index is changed by load, namely with the load rate of change size to this meteorological index under certain meteorological index best fit;
Consider the difference of each department meteorological factor and load, calculate its meteorological factor quantification index respectively by each department meteorological data, draw the quantification index value of each department further according to the load meteorology sensitivity size weighting of each department, namely
T=∑ Ti(ΔEmi/∑ΔEmi)
In formula, Ti is each department Meteorological Index, and Δ Emi is each department sensitivity numerical value;
So that the result of sensitive analysis more science is reliable, still uses multi-target analysis and select and the method on reserve section day off holiday is analyzed, the process of analysis being additionally added Thermal incubation effect so that analyze result more accurately reliable;
2) load forecasting model set:
21) based on the conventional forecast model of same type day, selection and the day to be predicted historical load day with type, analyze its Changing Pattern, arrange near to the order farthest away from day to be predicted, extract the Daily treatment cost with type day, day minimum load, per day load, and in this, as reference value, to this daily load curve standardization, adopt point-to-point multiple proportions method, equal lotus mode multiple proportions smoothing techniques, paddy lotus mode multiple proportions smoothing techniques, peak load mode multiple proportions smoothing techniques, equal lotus mode overlap smoothing techniques, paddy lotus mode overlap smoothing techniques, peak load mode overlap smoothing techniques, Linear regression is analyzed prediction;
22) forecast model is newly ceased, adopt up-to-date information that following daily load is predicted, to based on same type day model many algorithms improve, form new breath forecast model class algorithm: new cease point-to-point multiple proportions method, new cease equal lotus mode multiple proportions smoothing techniques, new breath paddy lotus mode multiple proportions smoothing techniques, new breath peak load mode multiple proportions smoothing techniques, new cease equal lotus mode overlap smoothing techniques, new breath paddy lotus mode overlap smoothing techniques, new breath peak load mode overlap smoothing techniques;
23) based on the forecast model of frequency domain decomposition method: electrically-based load has relatively strong periodicity, is used for time series frequency-domain analysis method and is predicted, and does following Fourier decomposition at Load Time Series P (t) specifying modeling time domain D-:
P ( t ) = a 0 + &Sigma; i = 1 N - 1 ( a i cos&omega; i t + b i sin&omega; i t ) ,
Wherein, N is the length of load sequence;
Character according to Fourier decomposition, the signal obtained after decomposition is mutually orthogonal directions, and load P (t) resolves into angular frequency in this way and isComponent, by suitable combination, and according to the periodic feature of load variations, P (t) is reconstructed following formula:
P (t)=a0+D(t)+W(t)+L(t)+H(t)
Wherein, diurnal periodicity component a0+ D (t) and week periodic component W (t) are the load components by fixed cycle change, and the algorithm of this model has low frequency average frequency domain component method, low frequency to smooth frequency domain components method, the relevant frequency domain components method of low frequency, the relevant frequency domain components method of surplus;
24) forecast model of meteorological factor is considered: the consideration meteorological factor forecast model based on mapping library set up for solving different factor dimension difference, the setting of different mappings value is carried out according to the actual weather in grid company location, pattern-recongnition method, similarity extrapolation, correlative factor matching method is adopted to carry out load prediction, not carrying out employing artificial neural network method when mapping library is arranged;
25) special forecast model festivals or holidays: the model built for country's legal major holiday or holiday, select and based on the conventional forecast model thought of same type day, it be predicted, concrete grammar have equal lotus mode multiple proportions smoothing techniques festivals or holidays, festivals or holidays paddy lotus mode multiple proportions smoothing techniques, festivals or holidays peak load mode multiple proportions smoothing techniques, pointwise festivals or holidays growth rate method。
5. short term power requirement analysis method according to claim 1, it is characterized in that, described forecast assessment mechanism refers to load index predictive value and actual value are carried out odds relatively, the practical effect of checking model, add up the whole network load prediction accuracy rate in many daily ranges, accuracy rate ranking and the contrast same period, realize the constant feedback to reasonable prediction effect, form prediction Closed-cycle effect, realized the lasting improvement of prediction effect by constantly experience accumulation。
CN201610016411.9A 2016-01-11 2016-01-11 Short-term electric power demand analysis method based on overall process technology improvement Pending CN105701570A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610016411.9A CN105701570A (en) 2016-01-11 2016-01-11 Short-term electric power demand analysis method based on overall process technology improvement

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610016411.9A CN105701570A (en) 2016-01-11 2016-01-11 Short-term electric power demand analysis method based on overall process technology improvement

Publications (1)

Publication Number Publication Date
CN105701570A true CN105701570A (en) 2016-06-22

Family

ID=56227267

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610016411.9A Pending CN105701570A (en) 2016-01-11 2016-01-11 Short-term electric power demand analysis method based on overall process technology improvement

Country Status (1)

Country Link
CN (1) CN105701570A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107067162A (en) * 2017-03-21 2017-08-18 浙江华立电力科技有限公司 Self study short-term load forecasting algorithm based on time domain and environmental characteristic
CN107292429A (en) * 2017-06-07 2017-10-24 上海欧睿供应链管理有限公司 A kind of Demand Forecast Model system of selection analyzed based on demand characteristics
CN107423836A (en) * 2017-04-07 2017-12-01 山东大学 Short-term load forecasting method based on sendible temperature
CN108205788A (en) * 2016-12-16 2018-06-26 国家电网公司 The multi objective electric load curve analysis system that a kind of outer source data crawls in real time
CN108932557A (en) * 2018-04-28 2018-12-04 云南电网有限责任公司临沧供电局 A kind of Short-term Load Forecasting Model based on temperature cumulative effect and grey relational grade

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108205788B (en) * 2016-12-16 2021-11-02 国家电网公司 Multi-index power load curve analysis system for crawling external source data in real time
CN108205788A (en) * 2016-12-16 2018-06-26 国家电网公司 The multi objective electric load curve analysis system that a kind of outer source data crawls in real time
CN107067162A (en) * 2017-03-21 2017-08-18 浙江华立电力科技有限公司 Self study short-term load forecasting algorithm based on time domain and environmental characteristic
CN107067162B (en) * 2017-03-21 2020-05-26 华立科技股份有限公司 Self-learning short-term load prediction method based on time domain and environmental characteristics
CN107423836A (en) * 2017-04-07 2017-12-01 山东大学 Short-term load forecasting method based on sendible temperature
CN107423836B (en) * 2017-04-07 2020-04-28 山东大学 Short-term load prediction method based on body sensing temperature
CN107292429A (en) * 2017-06-07 2017-10-24 上海欧睿供应链管理有限公司 A kind of Demand Forecast Model system of selection analyzed based on demand characteristics
CN108932557A (en) * 2018-04-28 2018-12-04 云南电网有限责任公司临沧供电局 A kind of Short-term Load Forecasting Model based on temperature cumulative effect and grey relational grade

Similar Documents

Publication Publication Date Title
CN105701570A (en) Short-term electric power demand analysis method based on overall process technology improvement
CN105069525B (en) Round-the-clock 96 Day Load Curve Forecastings and optimization update the system
CN105678404A (en) Micro-grid load prediction system and method based on electricity purchased on-line and dynamic correlation factor
Chikobvu et al. Regression-SARIMA modelling of daily peak electricity demand in South Africa
CN102426674B (en) Power system load prediction method based on Markov chain
US11043808B2 (en) Method for identifying pattern of load cycle
CN104636822A (en) Residential load prediction method of elman-based neural network
CN104598985A (en) Power load forecasting method
Phuangpornpitak et al. A study of load demand forecasting models in electric power system operation and planning
Qiuyu et al. Short-term load forecasting based on load decomposition and numerical weather forecast
CN102509173B (en) A kind of based on markovian power system load Accurate Prediction method
Perez Application of support vector machines and ANFIS to the short-term load forecasting
Smith Combination of forecasts in electricity demand prediction
CN104252647B (en) Electro-load forecast method based on anti-distance weighting interpolation method
CN107423836B (en) Short-term load prediction method based on body sensing temperature
CN108764547A (en) A kind of Mid-long Term Load curve generation method meeting Constraint
CN109858668B (en) Coordination prediction method for power load region in thunder and lightning climate
CN110033134A (en) A kind of short-term load forecasting algorithm of segmentation day by day considering meteorologic factor
KR100753141B1 (en) Prediction of temperature and humidity for estimation of cooling and heating load
CN108830415A (en) A kind of weather sensitive load power estimating method based on non-linear correlation model
CN109886567A (en) A kind of short-term load forecasting method considering sendible temperature and radiation intensity
JP3754267B2 (en) Water distribution prediction system
CN107067162A (en) Self study short-term load forecasting algorithm based on time domain and environmental characteristic
CN106529728A (en) Load zone prediction method with consideration of demand response measure
TWI498834B (en) Method for forecasting a short-term load of the next day for smart buildings

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20160622

WD01 Invention patent application deemed withdrawn after publication