CN117391734B - Power generation cost prediction method based on support vector machine regression model - Google Patents

Power generation cost prediction method based on support vector machine regression model Download PDF

Info

Publication number
CN117391734B
CN117391734B CN202311286295.9A CN202311286295A CN117391734B CN 117391734 B CN117391734 B CN 117391734B CN 202311286295 A CN202311286295 A CN 202311286295A CN 117391734 B CN117391734 B CN 117391734B
Authority
CN
China
Prior art keywords
cost
electricity
power generation
coal
unit
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.)
Active
Application number
CN202311286295.9A
Other languages
Chinese (zh)
Other versions
CN117391734A (en
Inventor
王辉
张依依
甘玮
韩金涛
王翔
薛泽彬
王晴
张子涵
费依蕃
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.)
North China Electric Power University
Original Assignee
North China Electric Power University
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 North China Electric Power University filed Critical North China Electric Power University
Priority to CN202311286295.9A priority Critical patent/CN117391734B/en
Publication of CN117391734A publication Critical patent/CN117391734A/en
Application granted granted Critical
Publication of CN117391734B publication Critical patent/CN117391734B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0206—Price or cost determination based on market factors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/27—Regression, e.g. linear or logistic regression
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Data Mining & Analysis (AREA)
  • Development Economics (AREA)
  • Evolutionary Computation (AREA)
  • Marketing (AREA)
  • Software Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Finance (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Human Resources & Organizations (AREA)
  • General Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Game Theory and Decision Science (AREA)
  • Mathematical Physics (AREA)
  • Tourism & Hospitality (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Biology (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Molecular Biology (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Primary Health Care (AREA)
  • Medical Informatics (AREA)

Abstract

The invention relates to a power generation cost prediction method based on a support vector machine regression model, and belongs to the technical field of coal-fired power generation. The method comprises the following steps: acquiring a data set of electricity generation cost of the degree obtained by calculation of coal-fired electricity generation parameters at each time point of history; acquiring corresponding boiler operation parameters of each time point of the history, and preprocessing to obtain an operation parameter data set; the coal-fired power generation parameters and the boiler operation parameters comprise the total coal quantity and the blending proportion of various coals; constructing a training set based on the operating parameter dataset and the power generation cost dataset; training the DA-SVM model by using a training set to obtain a trained DA-SVM model; the DA-SVM model is a support vector machine regression model based on dragonfly algorithm improvement; and predicting the power generation cost based on the current boiler operation parameters by using the trained DA-SVM model. The method can effectively and accurately predict the power generation cost based on the production history data of the coal-fired enterprises and guide the enterprises to adjust the production scheme.

Description

Power generation cost prediction method based on support vector machine regression model
Technical Field
The invention belongs to the technical field of coal-fired power generation, and particularly relates to a power generation cost prediction method based on a support vector machine regression model.
Background
With the continuous advancement and deepening of the reform of the electric power market, related enterprises acquire opportunities for surfing the internet to generate electricity by participating in bidding of the electric power market. In the process of coal-fired power generation production, different production schemes are adopted when the boiler operates, so that the cost change of each link of the coal-fired power generation production can be caused. The related enterprises at present mainly use 'post-accounting' in the aspect of power generation cost accounting, so that the calculation of the real-time power generation cost cannot be realized, and the cost possibly caused by the current production scheme cannot be predicted. Therefore, in order to effectively improve the competitiveness of enterprises, the enterprises are urgent in finding a reasonable and effective power generation cost prediction method.
Disclosure of Invention
In view of the above analysis, the present invention aims to provide a power generation cost prediction method based on a support vector machine regression model, which can effectively and accurately predict power generation cost based on production history data of a coal-fired enterprise. The method specifically comprises the following steps:
acquiring a data set of electricity generation cost of the degree obtained by calculation of coal-fired electricity generation parameters at each time point of history; acquiring boiler operation parameters corresponding to each time point to be predicted and each time point of the history to perform pretreatment, wherein the boiler operation parameters of each time point of the history after pretreatment form an operation parameter data set; the coal-fired power generation parameters and the boiler operation parameters comprise the total coal quantity and the blending proportion of various coals;
constructing a training set based on the operating parameter dataset and the power generation cost dataset;
training the DA-SVM model by using a training set to obtain a trained DA-SVM model; the DA-SVM model is a support vector machine regression model based on dragonfly algorithm improvement;
and predicting the power generation cost based on the preprocessed boiler operation parameters of each time point to be predicted by using the trained DA-SVM model.
Further, the training the DA-SVM model using the training set, and obtaining the trained DA-SVM model includes:
initializing iteration times, population scale of a dragonfly algorithm, position vector of each dragonfly individual in the population and step size vector of the dragonfly algorithm; wherein, a group of values of a parameter penalty factor c and a kernel function coefficient g of the SVM are combined to form a position vector of a dragonfly individual;
in each round of iteration: for each dragonfly individual, using SVM to obtain a corresponding prediction cost value of each sample based on the c and g values of the individual and the operation parameters of each sample in the training set, and calculating to obtain the adaptability value of the dragonfly individual based on all the prediction cost values and all the corresponding power generation cost; determining food positions, enemy positions and step vectors based on fitness values of all dragonfly individuals by using a dragonfly algorithm, and updating the dragonfly positions; the method comprises the steps of determining the individual position with the optimal fitness value in a current population as food, and determining the individual position with the worst fitness value in the current population as enemy;
judging whether a termination condition is reached, if not, continuing iteration; if yes, stopping iteration, and recording c and g values corresponding to the individual positions of the optimal fitness value as c and g values of the SVM, so as to obtain a trained DA-SVM model.
Further, the fitness function for calculating the fitness value is:
wherein n is the number of samples; p (P) i For each sample, the corresponding predicted cost value, yi is the corresponding electricity generation cost in each sample.
Further, the termination condition includes: the optimal fitness value is obtained by the iteration times or the dragonfly algorithm.
Further, the electricity generation cost comprises electricity coal cost and electricity operation cost; the electricity-measuring operation cost comprises electricity-measuring electricity generation fuel cost, electricity-measuring water cost, electricity-measuring station electricity cost, electricity-measuring maintenance cost and electricity-measuring discharge cost.
Further, the calculation method of the electricity-measuring coal cost comprises the following steps:
C coal ＝P 1 ·X sum ·X 1 +P 2 ·X sum ·X 2 +…+P i ·X sum ·X i …+P N ·X sum ·X N /(L f ×1000×T)；
wherein N represents the type of coal in the coal blending combustion scheme; x is X sum Representing the total coal amount; x is X i Represents the ratio of the ith coal in the total coal amount, andP i represents the price of the ith coal; l (L) f Is a power generation load; t is the power generation time.
Further, the electricity water cost comprises brine water production cost, circulating water production cost and reclaimed water cost, and the brine water production cost, the circulating water production cost and the reclaimed water cost are respectively calculated based on the flow of the demineralized water outlet pipe per minute, the flow of the circulating water replenishing water per minute and the flow of the reclaimed water main pipe per minute.
Further, the electricity discharge cost comprises limestone powder cost, environment-friendly discharge cost and electricity urea cost; the calculation method of the limestone powder cost comprises the following steps:
wherein,representing the SO of the raw flue gas 2 Reduced concentration in kg/Nm 3 ；C yq,in Represents the standard dry flow of the original flue gas, and the unit is m 3 /min；η ds The desulfurization efficiency is expressed in units of; />Represents CaCO 3 Molar mass (100 g/mol); />Representing SO 2 Molar amount (64 g/mol); r is (r) Ca/S Represents the calcium-sulfur ratio, the unit is; p (P) lime The purity of limestone powder is expressed in units of; p (P) shs The unit of limestone powder is yuan/ton; t represents the power generation time, and the unit is h.
Further, the emission environmental protection cost comprises electricity NO x Pollution discharge tax and electricity SO 2 The calculation methods of the blowdown tax and the electric smoke dust blowdown tax are as follows:
wherein,indicating the clean smoke NO x Concentration conversion in kg/Nm 3 ；C yq,out Represents the dry flow of the clean flue gas, and the unit is m 3 /min；/>Represents NO x Pollution equivalent value; the unit is m 3 /min；/>Represents NO x Unit price in yuan per kilogram; l (L) f The unit is MW for generating load;
indicating the SO of the clean flue gas 2 Concentration conversion in kg/Nm 3 ；/>Representing SO 2 Pollution equivalent value in m 3 /min；/>Representing SO 2 Unit price in yuan per kilogram;
C yc,out represents the conversion of the concentration of the smoke and dust of the clean smoke and the unit is kg/Nm 3 ；V yc Represents the equivalent value of smoke pollution, and the unit is m 3 /min；P yc The unit of smoke unit is yuan per kilogram.
Further, the calculation formula of the electricity maintenance cost is as follows:
C wh ＝(C cl +C jx )/P mf ；
wherein C is wh Representing the degree of electric maintenance cost, wherein the unit is meta/(kW.h); c (C) cl Representing the daily maintenance material cost of the current month, wherein the unit is unit; c (C) jx Representing the overhaul cost in the current month, wherein the unit is yuan; p (P) mf Indicating the total power generation amount in the current month.
The invention can realize at least one of the following beneficial effects:
calculating corresponding real-time electricity generation cost of each time point by collecting data of each production link of each time point of a coal-fired power generation enterprise; according to the real-time electricity generation cost of each time point of history, the corresponding relation between the boiler operation parameters and the electricity generation cost is searched by the training model, so that the corresponding electricity generation cost can be predicted efficiently and rapidly by using the model based on different production schemes of boiler operation, and on one hand, the adjustment of the production scheme can be guided according to the cost condition so as to achieve the aim of saving the cost; on the other hand, the power generation cost can be predicted according to the current production scheme, and the power plant is guided to formulate a reasonable online electronic price bidding strategy.
According to the invention, the Dragonfly Algorithm (DA) is adopted to select the optimal parameters of the Support Vector Machine (SVM), and as the dragonfly algorithm has stronger local searching capability and global searching capability, the dragonfly algorithm has good effect in parameter optimization, reduces the randomness of SVM prediction and improves the accuracy of prediction; in regression, the SVM selects a Gaussian radial basis function, so that the distribution structure of data can be described more accurately, and compared with other prediction models, the DA-SVM model adopted by the invention has more accurate prediction results, smaller errors and good generalization capability.
The accuracy of the calculation of the electricity generation cost is improved by considering the cost of each production link of the coal-fired power generation; in the calculation process, the minute-level calculation is realized for the cost of each link, so that the real-time performance of the calculation of the electricity generation cost is realized. The method comprises the steps of calculating to obtain minute-level electricity-measuring coal burning cost based on instantaneous coal quantity data of a coal machine; based on the treatment principle of sulfur dioxide and nitrogen oxide, the cost of pollutant treatment is calculated according to the flow rate of flue gas and the concentration value of pollutant by using a chemical formula, so that the accounting of the emission cost of 'minute' level can be realized; and calculating the electricity consumption water cost of the 'minute' level based on the water consumption of the circulating water, the desalted water and the reclaimed water pipeline flow value per minute of the thermal power enterprise. Compared with the post-calculation commonly adopted in the prior art, the minute-level electricity generation cost calculation is realized by comprehensively considering each production link and real-time data acquisition, the calculation result is more accurate, and the method is more beneficial to providing data support for decisions of a production scheme and bidding decisions for enterprises.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, like reference numerals being used to designate like parts throughout the drawings;
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a diagram of an iterative process for optimizing parameters c and g in the second embodiment;
FIG. 3 is a plot of the resulting selected contour of c, g in example two;
fig. 4 shows the prediction results of the DA-SVM model and other different models for predicting different samples in the second embodiment.
Detailed Description
Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form a part hereof, and together with the description serve to explain the principles of the invention, and are not intended to limit the scope of the invention.
Example 1
The invention discloses a power generation cost prediction method based on a support vector machine regression model, which comprises the following steps:
step S01, acquiring a data set of electricity generation cost of the electricity generation of the fire coal at each time point; acquiring boiler operation parameters corresponding to each time point to be predicted and each time point of the history to perform pretreatment, wherein the boiler operation parameters of each time point of the history after pretreatment form an operation parameter data set; the coal-fired power generation parameters and the boiler operation parameters comprise the total coal quantity and the blending proportion of various coals;
step S02, constructing a training set based on the operation parameter data set and the electricity generation cost data set;
s03, training the DA-SVM model by using a training set to obtain a trained DA-SVM model; the DA-SVM model is a support vector machine regression model based on dragonfly algorithm improvement;
and S04, predicting the power generation cost based on the preprocessed boiler operation parameters of each time point to be predicted by using the trained DA-SVM model.
Specifically, the boiler operation parameters in the step S01 include air supply temperature, condenser vacuum, water supply flow, total primary air quantity, total secondary air quantity, total air quantity of the boiler, hearth oxygen quantity, water supply temperature, total coal quantity and blending proportion of various coals; specifically, in this embodiment, blending ratios of various coals include: high-heat coal ratio, low-heat coal ratio and economic coal ratio.
Further, carrying out normalization pretreatment on the boiler operation parameters corresponding to each time point to be predicted and each historical time point to obtain an operation parameter data set, wherein the adopted normalization equation is as follows:the normalization processing of the data can reduce the influence of sample fluctuation and enhance the prediction performance.
It should be noted that in actual power generation, different boiler operation conditions may cause different costs of each production link of the power generation enterprise, that is, different boiler operation parameters correspond to different power generation costs. And the electricity generation cost is calculated and obtained based on the cost of each link of the electricity generation production.
The method of the embodiment obtains the real-time electricity generation cost of each corresponding time point of the coal-fired electricity generation enterprise by calculating the cost of each corresponding time point of each production link.
Specifically, in step S01, the electricity generation cost includes electricity fire coal cost and electricity operation cost; the electricity-measuring operation cost comprises electricity-measuring electricity generation fuel cost, electricity-measuring water cost, electricity-measuring station electricity cost, electricity-measuring maintenance cost and electricity-measuring discharge cost.
Further, the calculation method of the electricity-measuring coal cost and the related coal-burning power generation parameters are as follows:
C coal ＝P 1 ·X sum ·X 1 +P 2 ·X sum ·X 2 +…+P i ·X sum ·X i …+P N ·X sum ·X N /(L f ×1000×T)；
wherein N represents the type of coal in the coal blending combustion scheme; x is X sum The total coal amount is expressed in units of: t is; x is X i Represents the ratio of the ith coal in the total coal amount, andP i the price of the i-th coal is expressed in units of: meta/t; l (L) f The unit is MW for generating load; t is the power generation time, and the unit is h, and it is to be noted that T in all the formulas in the invention is the corresponding power generation time when each formula is calculated.
Further, in the present embodiment, n=3; the types of coal include high-volatility high-heat coal, high-volatility low-heat coal and economic coal; the calculation is performed based on the proportions of the respective kinds of coals.
Further, a calculation formula of the fuel cost of the power generation and related fuel power generation parameters are as follows:
C ry ＝P oil ×M oil /(L f ×1000×T)；
wherein C is ry The unit of the fuel cost of the power generation is meta/(kW.h); p (P) oil The fuel price is expressed in yuan/t; m is M oil The current fuel consumption is represented, and the unit is t (ton); l (L) f The unit is MW for the power generation load.
Further, the electricity water consumption cost comprises a brine water production cost, a circulating water production cost and a reclaimed water cost, wherein the brine water production cost, the circulating water production cost and the reclaimed water cost are respectively calculated based on the flow of a demineralized water outlet pipe per minute, the flow of circulating water replenishing water per minute and the flow of a reclaimed water main pipe per minute.
Further, the brine water production cost, the circulating water production cost and the reclaimed water cost, the calculation method and the related coal-fired power generation parameters (the coal-fired power generation parameters are water consumption parameters in coal-fired power generation) are respectively as follows:
C cys ＝G cys ×T×P cys /(L f ×1000×T)；
wherein C is cys The unit of the representative degree electric demineralized water production cost is Yuan/(kW.h); g cys The flow of the desalted water outlet mother pipe is expressed as t/h; p (P) cys The unit price of desalted water production is expressed, and the unit is yuan/t; l (L) f The unit is MW for generating load;
C xhs ＝G xhs ×T×P xhs /(L f ×1000×T)；
wherein C is xhs The unit of the representative degree electricity circulating water production cost is Yuan/(kW.h); g xhs The water supplementing flow of the circulating water is expressed, and the unit is t/h; p (P) xhs The unit price of the circulating water is represented, and the unit is yuan/t;
C zs ＝G zs ×P zs /(L f ×1000×T)；
wherein C is zs The unit of the water cost in the electric power is meta/(kW.h); g zs The flow of the water main pipe is expressed, and the unit is t/h; p (P) zs The unit price of the reclaimed water is expressed, and the unit is yuan/t.
Further, the electricity consumption cost of the power plant comprises the equipment electricity consumption cost of all production links of the thermal power enterprise, and a calculation formula and related coal-fired power generation parameters (the coal-fired power generation parameters are equipment electricity consumption parameters) are as follows:
C cyd ＝F e ×0.01×P e ；
wherein C is cyd The unit of the power consumption cost of the power station is meta/(kW.h); f (F) e Representing the power consumption of the plant; p (P) e The unit of the online electricity price is meta/(kW.h).
Further, the calculation formula of the electricity maintenance cost and the related coal-fired power generation parameters (equipment maintenance parameters) are as follows:
C wh ＝(C cl +C jx )/P mf ；
wherein C is wh Representing the degree of electric maintenance cost, wherein the unit is meta/(kW.h); c (C) cl Representing daily maintenance material fees (replacement parts) in units of elements in the month; c (C) jx Representing the overhaul cost in the current month, wherein the unit is yuan; p (P) mf Indicating the total power generation amount in the current month.
Further, the electricity discharge cost comprises limestone powder cost, environment-friendly discharge cost and electricity urea cost.
Specifically, the calculation method of limestone powder cost and related coal-fired power generation parameters (here, related enterprise emission parameters) are as follows:
wherein,representing the SO of the raw flue gas 2 Reduced concentration in kg/Nm 3 ；C yq,in Represents the standard dry flow of the original flue gas, and the unit is m 3 /min；η ds The desulfurization efficiency is expressed in units of; />Represents CaCO 3 Molar mass (100 g/mol); />Representing SO 2 Molar amount (64 g/mol); r is (r) Ca/S Represents the calcium-sulfur ratio, the unit is; p (P) lime The purity of limestone powder is expressed in units of; p (P) shs The unit of limestone powder is yuan/ton.
Further, the environmental protection cost of emission includes electricity NO x Pollution discharge tax and electricity SO 2 The pollution discharge tax and the electricity smoke dust pollution discharge tax are calculated by the following methods and related coal-fired power generation parameters (enterprise emission parameters):
wherein,indicating the clean smoke NO x Concentration conversion in kg/Nm 3 ；C yq,out Represents the dry flow of the clean flue gas, and the unit is m 3 /min；/>Represents NO x Pollution equivalent value; the unit is m 3 /min；/>Represents NO x Unit price in yuan per kilogram; l (L) f The unit is MW for generating load;
indicating the SO of the clean flue gas 2 Concentration conversion in kg/Nm 3 ；/>Representing SO 2 Pollution equivalent value in m 3 /min；/>Representing SO 2 Unit price in yuan per kilogram;
C yc,out represents the conversion of the concentration of the smoke and dust of the clean smoke and the unit is kg/Nm 3 ；V yc Represents the equivalent value of smoke pollution, and the unit is m 3 /min；P yc Represents smoke unit price in yuan per kilogram;
the flue gas standard dry flow is the flow obtained by converting the actually measured flue gas flow into the standard state (0 ℃ =273K, 101.325 Kpa), and the calculation method of the original flue gas standard dry flow and the net flue gas standard dry flow and the related coal-fired power generation parameters (enterprise emission parameters) are respectively as follows:
wherein C is yq,in Represents the standard dry flow of the original flue gas, and the unit is m 3 /min；S yyq Represents the cross-sectional area of the flue gas inlet, and the unit is m 2 ；V yyq The flow rate of the original flue gas is expressed, and the unit is m/s; f (F) a Atmospheric pressure is expressed in Pa; f (F) yyq The static pressure of the raw flue gas at the measuring point is expressed in Pa;the unit of the moisture content in the original flue gas is the volume percentage; t (T) yyq The temperature of the original flue gas is expressed in the unit of DEG C; c (C) yq,out Represents the dry flow of the clean flue gas, and the unit is m 3 /min；S jyq Represents the cross-sectional area of a flue gas outlet, and the unit is m 2 ；V jyq The flow rate of the clean flue gas is expressed in m/s; f (F) jyq The static pressure of the clean flue gas at the measuring point is expressed in Pa; />The unit of the water content volume percentage in the clean flue gas is shown as follows; t (T) jyq The net flue gas temperature is expressed in degrees celsius.
Further, the calculation method of the electricity-measuring urea cost and the related coal-fired power generation parameters (enterprise emission parameters) are as follows:
wherein C is ns Representing the electricity urea cost per unit cell/(kW.h);the ammonia amount (ammonia supply flow rate) is expressed in kg/h; p (P) ns The urea unit price is expressed in yuan/kg.
Further, the electricity-to-electricity running cost is the sum of the electricity-to-electricity power generation fuel cost, electricity-to-electricity water cost, electricity-to-electricity plant electricity cost, electricity-to-electricity maintenance cost and electricity-to-electricity discharge cost, and is expressed as: c (C) op ＝C ry +C ys +C cyd +C wh +C pf ；
Wherein C is ys Indicating the cost of electricity water, C ys ＝C cys +C xhs +C zs ；
C pf Indicating the electricity discharge cost, C pf ＝C shs +C hbs +C ns ；
C hbs Represents the environmental-protection cost of the emission,
further, the electricity generation cost is the sum of electricity coal cost and electricity operation cost, expressed as:
C＝C coal +C op 。
specifically, in step S02, the data in the operation parameter data set and the electricity generation cost data set are in one-to-one correspondence based on each point in time.
Specifically, in step S03, training the DA-SVM model using the training set, to obtain a trained DA-SVM model includes:
initializing iteration times, population scale of a dragonfly algorithm, position vector of each dragonfly individual in the population and step size vector of the dragonfly algorithm; the position vector of a dragonfly individual is formed by combining a group of values of a parameter penalty factor c and a kernel function coefficient g of the SVM, and the vector dimension is 2; wherein, the optimizing range of the parameters c and g is 2 -8 ～2 8 ；
In each round of iteration:
for each dragonfly individual, using SVM to obtain a corresponding prediction cost value of each sample based on the c and g values of the individual and the operation parameters of each sample in the training set, and calculating to obtain the adaptability value of the dragonfly individual based on all the prediction cost values and all the corresponding power generation cost;
determining the food position and the enemy position based on the fitness values of all dragonfly individuals by using a dragonfly algorithm, and updating the weight, the position vector and the step size vector of the dragonfly; the method comprises the steps of determining the individual position with the optimal fitness value in a current population as food, and determining the individual position with the worst fitness value in the current population as enemy; the optimal fitness value refers to the minimum value of all fitness values, and the worst fitness value refers to the maximum value of all fitness values;
at the completion of each iteration round:
judging whether a termination condition is reached, if not, continuing the next iteration; if yes, stopping iteration, and recording c and g values corresponding to individual positions of the optimal fitness value as c and g values of the SVM to obtain a trained DA-SVM model; wherein the termination condition includes: the optimal fitness value is obtained by the iteration times or the dragonfly algorithm.
Further, the fitness function for calculating the fitness value is:
wherein n is the number of samples; p (P) i For each sample, a corresponding predicted cost value, Y i The cost of power generation is measured for each sample.
Preferably, the SVM uses a Gaussian radial basis function.
Specifically, in step S04, based on the boiler operation parameters to be predicted at each time point after the normalization processing in step S01, the trained DA-SVM model is used to predict the corresponding power generation cost at each time point; specifically, each time point to be predicted refers to each time point selected in the current production time period of the cost to be predicted; optionally, the average value of the cost prediction results corresponding to each time point to be predicted is calculated as the cost prediction result of the current production time period.
The embodiment discloses a power generation cost prediction method based on a support vector machine regression model, which is characterized in that a corresponding relation between boiler operation parameters and power generation cost is searched by a training model according to real-time power generation cost of each time point of history, so that corresponding power generation cost can be predicted efficiently and rapidly by using different production schemes of the model based on boiler operation, and on one hand, the adjustment of the production scheme can be guided according to cost conditions so as to achieve the aim of saving cost; on the other hand, the power generation cost can be predicted according to the current production scheme, and the power plant is guided to formulate a reasonable online electronic price bidding strategy.
The optimal parameters of the Support Vector Machine (SVM) are selected by adopting a Dragonfly Algorithm (DA), and the dragonfly algorithm has good effect in parameter optimization because of strong local searching capability and global searching capability, so that the randomness of SVM prediction is reduced, and the accuracy of prediction is improved.
In the process of calculating the historical power generation cost, the cost of each production link of the coal-fired power generation is considered, the minute-level calculation is realized on the cost of each link, the instantaneity and the accuracy of the calculation of the power generation cost are realized, and compared with the post calculation commonly adopted in the prior art, the real-time acquisition of each production link and the data is comprehensively considered, the calculation result is more accurate, and the data support is more beneficial to the decision of the production scheme and the bid decision provided for enterprises.
Example two
The invention discloses a power generation cost prediction method based on a support vector machine regression model, which comprises the following steps:
step S11, S01 same as the first embodiment, specifically, selecting a fire coal power generation parameter value of a minute-level time point in the last two months of history, and calculating to obtain corresponding minute-level electricity generation cost data, so as to form an electricity generation cost data set.
And step S12, constructing a training set based on the operation parameter data set and the electricity generation cost data set.
S13, training the DA-SVM model by using a training set to obtain a trained DA-SVM model; the DA-SVM model is a support vector machine regression model based on dragonfly algorithm improvement;
specifically, the training iteration number initialization value of the DA-SVM model is 100; the population size is 50, namely the number of individuals in the dragonfly population is 50; c. g the optimizing range of the two parameters is 2 -8 ～2 8 The method comprises the steps of carrying out a first treatment on the surface of the FIGS. 2 and 4 show the seeking of the model of the present embodimentAnd (5) optimizing the process. Wherein, FIG. 2 is a optimizing iterative process of parameters c, g; FIG. 3 is a plot of the resulting selected contour for c, g (the data in the contour is identified as 10 of the actual MSE 6 A multiple); when the fitness function reaches a minimum, c=0.3100, g= 0.2867.
And S14, predicting the power generation cost based on the boiler operation parameters at each time point to be predicted by using the trained DA-SVM model.
Specifically, in this embodiment, the boiler operation parameters at each time point in the current week of production are selected, and the cost prediction is performed based on the current week of production scheme.
Specifically, in this embodiment, in order to verify the performance of the DA-SVM model, a Random Forest (RF) model and a BP neural network (Back Propagation Neural Network, BP) model are used to predict the unit operation cost of the same sample set, respectively. And respectively comparing the actual electric operation cost with the prediction results of the three models. Wherein the number of the initialized parameter trees in the RF model is 50 (n_identifiers=50), and the other is a default value; the learning rate of the initialization parameter in the BP model was 0.02 (learning rate=0.02), the training number was 300 (epochs=300), and the number of hidden layer nodes was 100 (Number of hidden layers =100). Fig. 4 shows the prediction results of the prediction of different samples using different models, and table 1 shows the comparison of the evaluation indexes of the prediction results of different models.
Table 1, comparison of evaluation index of prediction result
Wherein, the expressions of each evaluation index are respectively:
mean square error
Error of allSquare root
Average absolute error
Average absolute percentage error
As can be seen from the comparison of Table 1, the evaluation index results of the DA-SVM model are: MSE is 3.84×10 -6 RMSE 0.00196, mae 0.00163, mape 0.02657. The 3 error indexes of the DA-SVM model are the smallest in all methods, and the DA-SVM shows higher prediction accuracy. The prediction advantage is embodied. The DA-SVM model can realize accurate prediction of the electricity generation cost, and has good popularization effect.
It should be noted that, the above embodiments are based on the same inventive concept, and the description is not repeated, and the description may be referred to each other.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention.

Claims (4)

1. The power generation cost prediction method based on the support vector machine regression model is characterized by comprising the following steps of:
acquiring a data set of electricity generation cost of the degree obtained by calculation of coal-fired electricity generation parameters at each time point of history;
the electricity generation cost comprises electricity coal cost and electricity operation cost;
the electricity-measuring operation cost comprises electricity-measuring electricity generation fuel cost, electricity-measuring water cost, electricity-measuring station electricity cost, electricity-measuring maintenance cost and electricity-measuring discharge cost;
the calculation method of the electricity-measuring coal cost comprises the following steps:
C coal ＝P 1 ·X sum ·X 1 +P 2 ·X sum ·X 2 +…+P i ·X sum ·X i …+P N ·X sum ·X N /(L f ×1000×T)；
wherein N represents the type of coal in the coal blending combustion scheme; x is X sum Representing the total coal amount; x is X i Represents the ratio of the ith coal in the total coal amount, andP i represents the price of the ith coal; l (L) f Is a power generation load; t is the power generation time;
the electricity water cost comprises brine water production cost, circulating water production cost and reclaimed water cost, and the brine water production cost, the circulating water production cost and the reclaimed water cost are respectively calculated based on the flow of a demineralized water outlet pipe per minute, the flow of circulating water replenishing water per minute and the flow of a reclaimed water main pipe per minute;
the electricity-metering discharge cost comprises limestone powder cost, environment-friendly discharge cost and electricity-metering urea cost; the calculation method of the limestone powder cost comprises the following steps:
wherein,representing the SO of the raw flue gas 2 Reduced concentration in kg/Nm 3 ；C yq,in Represents the standard dry flow of the original flue gas, and the unit is m 3 /min；η ds The desulfurization efficiency is expressed in units of; />Represents CaCO 3 Molar mass (100 g/mol); />Representing SO 2 Molar amount (64 g/mol); r is (r) Ca/S Represents the calcium-sulfur ratio, the unit is; p (P) lime The purity of limestone powder is expressed in units of; p (P) shs The unit of limestone powder is yuan/ton; t represents power generation time, and the unit is h;
the emission environmental protection cost comprises electricity NO x Pollution discharge tax and electricity SO 2 The calculation methods of the blowdown tax and the electric smoke dust blowdown tax are as follows:
wherein,indicating the clean smoke NO x Concentration conversion in kg/Nm 3 ；C yq,out Represents the dry flow of the clean flue gas, and the unit is m 3 /min；/>Represents NO x Pollution equivalent value; the unit is m 3 /min；/>Represents NO x Unit price in yuan per kilogram; l (L) f The unit is MW for generating load;
indicating the SO of the clean flue gas 2 Concentration conversion in kg/Nm 3 ；/>Representing SO 2 Pollution equivalent value in m 3 /min；Representing SO 2 Unit price in yuan per kilogram;
C yc,out represents the conversion of the concentration of the smoke and dust of the clean smoke and the unit is kg/Nm 3 ；V yc Represents the equivalent value of smoke pollution, and the unit is m 3 /min；P yc Represents smoke unit price in yuan per kilogram;
the calculation formula of the electricity maintenance cost is as follows:
C wh ＝(C cl +C jx )/P mf ；
wherein C is wh Representing the degree of electric maintenance cost, wherein the unit is meta/(kW.h); c (C) cl Representing the daily maintenance material cost of the current month, wherein the unit is unit; c (C) jx Representing the overhaul cost in the current month, wherein the unit is yuan; p (P) mf Representing the total power generation amount in the current month;
acquiring boiler operation parameters corresponding to each time point to be predicted and each time point of the history to perform pretreatment, wherein the boiler operation parameters of each time point of the history after pretreatment form an operation parameter data set; the coal-fired power generation parameters and the boiler operation parameters comprise the total coal quantity and the blending proportion of various coals;
constructing a training set based on the operating parameter data set and the electrical power generation cost data set;
training the DA-SVM model by using a training set to obtain a trained DA-SVM model; the DA-SVM model is a support vector machine regression model based on dragonfly algorithm improvement;
and predicting the power generation cost based on the preprocessed boiler operation parameters of each time point to be predicted by using the trained DA-SVM model.
2. The method of claim 1, wherein training the DA-SVM model using the training set to obtain a trained DA-SVM model comprises:
initializing iteration times, population scale of a dragonfly algorithm, position vector of each dragonfly individual in the population and step size vector of the dragonfly algorithm; wherein, a group of values of a parameter penalty factor c and a kernel function coefficient g of the SVM are combined to form a position vector of a dragonfly individual;
in each round of iteration: for each dragonfly individual, using SVM to obtain a corresponding prediction cost value of each sample based on the c and g values of the individual and the operation parameters of each sample in the training set, and calculating to obtain the adaptability value of the dragonfly individual based on all the prediction cost values and all the corresponding power generation cost; determining the food position and the enemy position based on the fitness values of all dragonfly individuals by using a dragonfly algorithm, and updating the weight, the position vector and the step size vector of the dragonfly; the method comprises the steps of determining the individual position with the optimal fitness value in a current population as food, and determining the individual position with the worst fitness value in the current population as enemy;
judging whether a termination condition is reached, if not, continuing iteration; if yes, stopping iteration, and recording c and g values corresponding to the individual positions of the optimal fitness value as c and g values of the SVM, so as to obtain a trained DA-SVM model.
3. The power generation cost prediction method according to claim 2, wherein the fitness function that calculates the fitness value is:
wherein n is the number of samples; p (P) i For each sample, a corresponding predicted cost value, Y i The cost of power generation is measured for each sample.
4. A power generation cost prediction method according to claim 3, wherein the termination condition includes: the optimal fitness value is obtained by the iteration times or the dragonfly algorithm.
CN202311286295.9A 2023-10-07 2023-10-07 Power generation cost prediction method based on support vector machine regression model Active CN117391734B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311286295.9A CN117391734B (en) 2023-10-07 2023-10-07 Power generation cost prediction method based on support vector machine regression model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311286295.9A CN117391734B (en) 2023-10-07 2023-10-07 Power generation cost prediction method based on support vector machine regression model

Publications (2)

Publication Number Publication Date
CN117391734A CN117391734A (en) 2024-01-12
CN117391734B true CN117391734B (en) 2024-03-26

Family

ID=89436534

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311286295.9A Active CN117391734B (en) 2023-10-07 2023-10-07 Power generation cost prediction method based on support vector machine regression model

Country Status (1)

Country Link
CN (1) CN117391734B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102778538A (en) * 2012-07-06 2012-11-14 广东电网公司电力科学研究院 Soft measuring method based on improved SVM (Support Vector Machine) for measuring boiler unburned carbon content in fly ash
CN107133460A (en) * 2017-04-26 2017-09-05 中国能源建设集团广东省电力设计研究院有限公司 A kind of online dynamic prediction method of boiler flyash carbon content
CN112598167A (en) * 2020-12-17 2021-04-02 上海电力大学 Power station boiler NO based on dragonfly algorithm and fast learning netxEmission amount prediction method
WO2023279636A1 (en) * 2021-07-09 2023-01-12 南京航空航天大学 Method for predicting material demands in assembly workshop
CN116307075A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Intelligent algorithm-based electricity-measuring coal-burning cost optimization method and system

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110443002A (en) * 2019-08-16 2019-11-12 中国水利水电科学研究院 A kind of Deformation of Steep Slopes prediction technique and system

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102778538A (en) * 2012-07-06 2012-11-14 广东电网公司电力科学研究院 Soft measuring method based on improved SVM (Support Vector Machine) for measuring boiler unburned carbon content in fly ash
CN107133460A (en) * 2017-04-26 2017-09-05 中国能源建设集团广东省电力设计研究院有限公司 A kind of online dynamic prediction method of boiler flyash carbon content
CN112598167A (en) * 2020-12-17 2021-04-02 上海电力大学 Power station boiler NO based on dragonfly algorithm and fast learning netxEmission amount prediction method
WO2023279636A1 (en) * 2021-07-09 2023-01-12 南京航空航天大学 Method for predicting material demands in assembly workshop
CN116307075A (en) * 2023-02-01 2023-06-23 华能国际电力股份有限公司上海石洞口第二电厂 Intelligent algorithm-based electricity-measuring coal-burning cost optimization method and system

Also Published As

Publication number Publication date
CN117391734A (en) 2024-01-12

Similar Documents

Publication Publication Date Title
CN113887827B (en) Coal blending and burning optimization decision method based on real-time carbon emission monitoring of thermal power generating unit
Zhao et al. Scenario analysis of the carbon pricing policy in China’s power sector through 2050: Based on an improved CGE model
CN109345012B (en) Park energy Internet operation optimization method based on comprehensive evaluation indexes
CN109741103B (en) Short-term multi-objective double-layer optimized scheduling method
CN111985702A (en) Park level comprehensive energy system optimization method considering electric energy substitution effect
CN110738530A (en) user side-oriented electric energy substitution comprehensive benefit analysis method
CN112836429A (en) Multi-objective optimization coal blending method based on coal quality prediction
CN115018230A (en) Low-carbon robust economic optimization operation method of comprehensive energy system considering emission reduction cost
CN106779361B (en) The multi-objective optimization design of power method of grid-connected mixing renewable energy system
CN111210079A (en) Operation optimization method and system for distributed energy virtual power plant
Han et al. Coal-fired power plant CCUS project comprehensive benefit evaluation and forecasting model study
CN115423282A (en) Electricity-hydrogen-storage integrated energy network multi-objective optimization scheduling model based on reward and punishment stepped carbon transaction
CN113779792A (en) Affine-based comprehensive energy system optimal configuration method
CN111949936A (en) Method and system for evaluating comprehensive energy system of suburb
CN114154744A (en) Capacity expansion planning method and device of comprehensive energy system and electronic equipment
CN115936220A (en) Garden comprehensive energy system planning and operation optimization method and terminal
Alam et al. Simulated energy scenarios of the power sector in Bangladesh
CN117391734B (en) Power generation cost prediction method based on support vector machine regression model
CN112653180B (en) Wind-fire-storage combined system environment economic dispatching method and system
CN112381268B (en) Short-term coal burning cost prediction method and system for electric power spot market
CN115496378A (en) Power system economic dispatching method taking wind energy emission reduction benefits into account
CN115438932A (en) Optimized scheduling and immune particle swarm solving method for comprehensive energy distribution network
CN115496336A (en) Thermal power generating unit benefit evaluation method under peak regulation auxiliary service market mechanism
CN114925892A (en) Water-electricity-to-gas combined medium-and-long-term wind-water-fire generating capacity double-layer planning method
CN113450186A (en) Bidding strategy applicable to monthly centralized market with participation of coupled system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant