CN113159835B - Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment - Google Patents
Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN113159835B CN113159835B CN202110374116.1A CN202110374116A CN113159835B CN 113159835 B CN113159835 B CN 113159835B CN 202110374116 A CN202110374116 A CN 202110374116A CN 113159835 B CN113159835 B CN 113159835B
- Authority
- CN
- China
- Prior art keywords
- price
- quotation
- unit
- day
- model
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S50/00—Market activities related to the operation of systems integrating technologies related to power network operation or related to communication or information technologies
- Y04S50/14—Marketing, i.e. market research and analysis, surveying, promotions, advertising, buyer profiling, customer management or rewards
Abstract
The embodiment of the application discloses a power generation side electricity price quotation method and device based on artificial intelligence, a storage medium and electronic equipment, and belongs to the field of electric power metering.
Description
Technical Field
The application relates to the field of electric power metering, in particular to a method and a device for electricity price quotation on a power generation side based on artificial intelligence, a storage medium and electronic equipment.
Background
In order to guarantee safe and stable operation of a power grid, promote consumption of clean energy and meet real-time balance of supply and demand of a system, 8 provinces such as Guangdong, shanxi and the like are selected by national development and transformation committee as the first trial points of domestic electric power spot market transactions. The commercial power generation side of the domestic market (namely spot goods) mainly adopts the mode of subsection volume report quotation and no quotation of the user side, and a time-of-use power generation output curve and a node marginal power price are obtained through optimization calculation. On this background, how to maximize the generating profit by designing a reasonable quotation curve according to the bidding behaviors of other participants and the operation condition of the power system is a hot spot of current research.
Disclosure of Invention
The embodiment of the application provides a method and a device for power generation side price quotation based on artificial intelligence, a storage medium and electronic equipment, and the problem of maximization of power generation profit can be solved. The technical scheme is as follows:
in a first aspect, an embodiment of the present application provides an artificial intelligence-based electricity price quotation method on a power generation side, where the method includes:
selecting relevant characteristics from the potential influence characteristics of the electricity price;
training the sample data of the relevant features by using an integrated learning framework based on a decision tree to obtain a feature training model;
inputting data to be predicted into the feature training model to obtain a predicted value for determining the marginal price of the market node at the day before;
processing the determined predicted value of the marginal price of the day-ahead market node by using a quantile regression model to obtain the uncertain predicted value of the marginal price of the day-ahead market node;
calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node;
randomly generating N price scenes; wherein N is an integer greater than 1, and the probability of each price scene is 1/N;
and constructing a quotation decision model under each price scene, and determining the volume benefit and quotation scheme under different price scenes by using the constructed quotation decision model.
In a second aspect, an embodiment of the present application provides an artificial intelligence-based electricity price quotation device on a power generation side, including:
a selection unit for selecting a relevant feature among the potentially influencing features of the electricity prices;
the training unit is used for training the sample data of the relevant features by utilizing an integrated learning framework based on a decision tree to obtain a feature training model;
the prediction unit is used for inputting data to be predicted into the feature training model to obtain a predicted value for determining the marginal price of the market node at the present;
the prediction unit is also used for processing the day-ahead market node marginal price determination prediction value by utilizing a quantile regression model to obtain a day-ahead market node marginal price uncertainty prediction value; calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node;
a determining unit for randomly generating N price scenes; wherein N is an integer greater than 1, and the probability of each price scene is 1/N;
and constructing a quotation decision model in each price scene, and determining the volume benefit and quotation scheme in different price scenes by using the constructed quotation decision model.
In a third aspect, embodiments of the present application provide a computer storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the above-mentioned method steps.
In a fourth aspect, an embodiment of the present application provides an electronic device, which may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the above-mentioned method steps.
The beneficial effects brought by the technical scheme provided by some embodiments of the application at least comprise:
the method comprises the steps of firstly, predicting node marginal electricity prices by using an integrated learning framework based on a decision tree of an artificial intelligence technology, obtaining electricity price distribution of different confidence intervals according to a quantile regression method, then, constructing a quotation decision model by combining unit physical constraints and price scenes, solving optimal parameters of the quotation decision model by combining the artificial intelligence technology, improving optimizing speed by introducing a momentum theory, and finally, establishing different quantity cost schemes for comparison according to different electricity price scenes, thereby providing an effective evaluation basis for reasonable quotation and risk management of a power generation side.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a network structure diagram provided in an embodiment of the present application;
fig. 2 is a schematic flowchart of an artificial intelligence-based electricity price quotation method on a power generation side according to an embodiment of the present application;
fig. 3 is another schematic flow chart of an artificial intelligence-based electricity price quotation method on a power generation side according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of an apparatus provided in an embodiment of the present application;
fig. 5 is a schematic structural diagram of an apparatus according to an embodiment of the present disclosure.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
The following description refers to the accompanying drawings in which the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the application, as detailed in the appended claims.
In the description of the present application, it is to be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. The specific meaning of the above terms in the present application can be understood in a specific case by those of ordinary skill in the art. Further, in the description of the present application, "a plurality" means two or more unless otherwise specified. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
Referring to fig. 1, a network architecture diagram provided in an embodiment of the present application is a network architecture diagram, where the network architecture of the present application includes: electronic device 11 and server 12.
The electronic equipment 11 is deployed at a power generation enterprise side, the server 12 is deployed at a power trading center, and the electronic equipment 11 is used for collecting power trading data on the server 12 and generating an offer curve by using the power trading data and power generation data detected at a power plant so as to maximize enterprise profits. The electronic device 11 communicates with the server 12 by using a wired communication method (such as optical fiber, twisted pair or power line) or a wireless communication method (such as bluetooth, microwave or radio frequency). The number of the servers 12 may be one or more, and the number of the electronic devices 11 may be one or more.
In the related art, power generation enterprises maximize power generation profits by designing reasonable quotation curves according to bidding behaviors of other participants and operation conditions of a power system. Generally, methods and systems for establishing day-ahead market bidding strategies are mainly classified into three main categories: the first type is a marginal cost-based quotation strategy; the second type is a quotation strategy based on a game theory; the third category is data-driven based bidding strategies.
And in the first category, quotation is carried out by adopting a marginal profit maximization principle: if the unit quoted price is higher than the node marginal electricity price, the unit has loss; and if the unit quoted price is lower than the node marginal electricity price, the unit cannot bid. However, this method has the following disadvantages: 1) If the unit is a pricing unit, which is equivalent to the price quoted to influence the marginal electricity price of the node, the assumption of marginal profit maximization is not established; 2) Because the marginal electricity price number of the nodes in the market at the day before is not equal to the number of the quotation segments, the conclusion of marginal profit maximization is not established; 3) Because the unit has the minimum startup time, if the marginal cost in a certain time period is lower than the marginal price of electricity of the node, the unit cannot be normally shut down; 4) Because the unit is limited by climbing, if the marginal price of the node is changed greatly, the unit cannot output power normally; 5) Marginal cost is related to real-time heat rate and real-time fuel price, but it is difficult to accurately evaluate them by various methods at present. And in the second category, according to the quotation strategy and the power grid operating environment of the opposite side, the clearing is simulated by adopting a safety constraint unit combination and a safety constraint economic dispatching, and the quotation strategy of the opposite side is converted through a clearing result so as to achieve the purpose of winning. Due to the strong theoretical logic, the quotation strategy based on the game theory is widely applied to quotation and market analysis of power generation enterprises. However, this method has the following disadvantages: 1) The quotation strategy of the other party cannot be obtained; 2) All marginal conditions of the power grid operation environment cannot be acquired; 3) The safety constraint unit combination and the safety constraint economic dispatching can not be consistent with an optimization model of a power grid dispatching center; 4) Because the optimization model is too complex, the optimization equilibrium solution is not provided. In summary, the first category and the second category are difficult to realize the price quotation decision, so that much attention is not needed.
And thirdly, forecasting the marginal electricity price of the market node at the day before by adopting a statistical regression method so as to formulate reasonable quotation. However, the current method mainly has the following defects: 1) The statistical regression method mainly focuses on machine learning of a shallow structure, but since the shallow machine learning model is easy to over-fit and disappear in gradient, the statistical regression method is limited when the node marginal electricity price prediction problem is processed; 2) Uncertainty of node marginal electricity price prediction is not considered; 3) A reasonable price model is constructed without consideration; 4) And a reasonable solution algorithm considering the quotation model is lacked.
In order to solve the above problems, the present application makes the following 6 improvements:
1. and (4) selecting the characteristics. The traditional characteristic Selection methods such as Pearson correlation, chi-square inspection, information gain and the like are abandoned, and a Lasso algorithm (LeastAbolute Shrinkage and Selection Operator, lasso) is introduced to carry out correlation analysis on the electricity price correlation characteristic, so that the characteristic dimension reduction can be effectively reduced.
2. And (5) feature training. Shallow machine learning methods such as support vector regression, decision trees, K neighborhood and neural networks are abandoned, and the improved lightweight gradient elevator is adopted to perform feature training on the marginal electricity price of the market node in the future, so that the prediction precision can be effectively improved.
3. And (5) interval prediction. Traditional interval price prediction methods such as Gaussian distribution, logistic distribution, t distribution and the like are abandoned, a quantile regression model is introduced to conduct interval prediction on day-ahead market node marginal price data of point prediction, and uncertainty of day-ahead market node marginal prices can be effectively reduced.
4. And constructing a quotation model. The game theory-based power generation side quotation strategy and the marginal cost-based power generation side quotation strategy are abandoned, the price prediction-based day-ahead market power generation side quotation decision model is provided, and the power generation benefits of the day-ahead market power generation side can be effectively improved.
5. And (5) solving an algorithm. Solving methods such as a particle swarm algorithm, a genetic algorithm, a traditional firefly algorithm and the like are abandoned, the momentum firefly algorithm is proposed to solve the quotation decision model, and the momentum firefly algorithm can avoid the problems of local optimization, low convergence speed and the like by introducing momentum parameters.
6. And (5) comparing the schemes. A quotation strategy of a single scheme is abandoned, multi-dimensional comparison is realized through different price scenes, cost and cost are calculated, and uncertain risk control of market quotation in the future can be effectively improved.
Referring to fig. 2, a schematic flow chart of a power generation-side electricity price quotation method based on artificial intelligence according to an embodiment of the present application is shown based on the network architecture of fig. 1. As shown in fig. 2, the method of the embodiment of the present application may include the steps of:
s201, selecting relevant characteristics from the potential influence characteristics of the electricity price.
The potential invisible features are parameters having potential influence on the electricity price, and the purpose of feature selection is to select related features closer to the electricity price from a plurality of candidate potential influence features so as to reduce the computation amount of model training.
S202, training the sample data of the relevant features by using an integrated learning framework based on a decision tree to obtain a feature training model.
The sample data comprises parameter values of the relevant features, and the sample data is marked data.
And S203, inputting the data to be predicted into the feature training model to obtain a predicted value for determining the marginal price of the market node in the day ahead.
The data to be predicted is a parameter value which is acquired by the trading center and contains the relevant characteristics.
And S204, processing the determined predicted value of the marginal price of the day-ahead market node by using a quantile regression model to obtain the uncertain predicted value of the marginal price of the day-ahead market node.
And S205, calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node.
And S206, randomly generating N price scenes.
Wherein N is an integer greater than 1, and the probability of each price scenario is 1/N.
S207, constructing a quotation decision model in each price scene, and determining the volume profit and the quotation scheme in different price scenes by using the constructed quotation decision model.
The quotation decision model can be a quotation curve and represents the relation between time and price, the volume cost represents the yield cost profit, and the maximum profit can be calculated by analyzing the relation among the production cost, the sales profit and the product quantity. The quotation scheme represents a quotation rule of the power plant based on a quotation curve.
According to the embodiment, firstly, an integrated learning framework based on a decision tree of an artificial intelligence technology is used for predicting node marginal electricity prices, electricity price distribution of different confidence intervals is obtained according to a quantile regression method, then a quotation decision model is constructed by combining unit physical constraints and price scenes, optimal parameters of the quotation decision model are solved by combining the artificial intelligence technology, optimizing speed is improved by introducing a momentum theory, and finally different amount cost schemes are established for comparison according to different electricity price scenes, so that an effective evaluation basis is provided for reasonable quotation and risk management of a power generation side.
Referring to fig. 3, another schematic flow chart of an artificial intelligence-based electricity price quotation method on a power generation side according to an embodiment of the present application is provided, in the embodiment of the present application, the method includes the following steps:
s301, selecting related features from the multiple potential influence features of the electricity price by using a lasso algorithm.
The method comprises the steps of firstly, carrying out feature selection on a plurality of potential influence features of the electricity price, and then carrying out feature training on related features by adopting an integrated learning method so as to realize point prediction of the marginal price of a market node in the future, namely deterministic prediction.
The purpose of feature Selection is to select influence factors closely related to electricity price, and a Lasso algorithm (Least Absolute Shrinkage and Selection Operator, lasso) is introduced to carry out correlation analysis on a plurality of potential influence features, wherein the specific formula is as follows:
in the formula (1), x tj The characteristics of the j-th time-sharing related potential influence at the t-th time generally comprise historical day-ahead/real-time market price, provincial load, outgoing/incoming call, class-A unit load, load bidding space and the like. r and beta j Are regression coefficients.
If the selection coefficient is smaller, the time-sharing irrelevant characteristic can be eliminated; t is the number of predicted time periods, argmin represents the minimum, and p represents the number of potentially influencing features.
And S302, training the sample data of the relevant characteristics by using a lightweight gradient elevator to obtain a characteristic training model.
Wherein, the purpose of the feature training is to extract the nonlinear/linear relation between the relevant features and the electricity price. The Light Gradient Boosting Machine (LGBM) is an integrated learning framework based on a decision tree, and compared with integrated learning methods such as a Boosting tree, a Gradient Boosting tree, and extreme Gradient Boosting, the LGBM grows layer by layer and limits the depth of the tree by the maximum depth and the number of leaves of an additional parameter tree, so that the LGBM has the advantages of high training speed, high convergence accuracy and the like. The sample data for price prediction in the initial stage of the power market is rare, and the LGBM needs sufficient samples to ensure data fitting, so that the defects of poor stability, overfitting and the like are caused. Aiming at the problems, the L2 regularization term is introduced into a loss function of the LGBM, so that the problems of overfitting and poor stability are solved, and the specific formula is as follows:
in the formula (2), J is the true value y t An error function with the predicted value of the M-1 model, here a logarithmic loss function; t is m-1 Fitting values (training values) composed of previous M-1 tree models; z is a radical of formula tj Selecting a jth correlation feature in the tth time series for the feature; a. The m-1 Is the parameter (xi) of the top M-1 tree model 1 ，ξ 2 ，…，ξ M-1 ) (ii) a λ is the regularization term coefficient.
And S303, inputting the data to be predicted into the feature training model to obtain a day-ahead market node marginal price determination predicted value.
The method comprises the following steps of substituting data to be predicted of relevant characteristics newly issued by a trading center into a characteristic training model to carry out deterministic prediction (point prediction) on the marginal price of the market node in the day-ahead, wherein the specific formula is as follows
In the formula, p t * Determining a predicted value for the marginal price of the market node at the moment t; z is a radical of tj * Is the relevant feature of the jth at time t.
And S304, processing the day-ahead market node marginal price determined predicted value by using a quantile regression model to obtain the day-ahead market node marginal price determined predicted value.
Where the uncertainty risk control of market quotes at a day-ahead is not favoured, considering that the error of deterministic price forecasts (point forecasts) is always present. According to the method, uncertainty regression is firstly carried out through a quantile regression model so as to realize uncertainty prediction of the marginal price of the market node in the day ahead.
Step 2.1: the uncertainty regression refers to the probability statistical analysis of the historical predicted value and the predicted true value in different confidence intervals. The quantile regression as an uncertainty regression method can effectively solve the probability statistics problem of price prediction, and the specific formula is as follows:
in equation 4, ρ τ Is a piecewise linear loss function at the tau quantile; beta is a 1 ， τ And beta 0 ， τ Are the regression coefficients at the τ quantiles, respectively; r is a radical of hydrogen t Is an error value, p t Is A m-1 。
Step 2.2: substituting the predicted value determined by the marginal price of the day-ahead market node into the quantile regression model to predict the uncertainty of the marginal price of the day-ahead market node, wherein the specific formula is as follows:
in formula 5, p t * The uncertain predicted value of the marginal price of the market node at the day before is obtained; y is τ Predicted values at τ quantile.
And S305, calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node.
S306, randomly generating N price scenes by adopting a Monte Carlo simulation method.
On the basis of uncertain price prediction, firstly a certain confidence interval CI (namely an upper quantile and a lower quantile are determined) is given, an interval prediction value is obtained through a quantile regression model, then N price scenes are randomly generated by adopting a Monte Carlo simulation method, N is an integer larger than 1, and the probability of each scene is assumed to be 1/N.
And S307, constructing a quotation decision model in each price scene.
On the basis of N price scenes, the method for constructing the auxiliary decision quotation model of the day-ahead market power generation side comprises the following steps: the method takes the expected income of the maximum day-ahead market power generation side as a target function, takes unit climbing, medium-long term contracts, declaration rules and the like as constraint conditions, and takes a quotation curve (sectional quotation) of the day-ahead market power generation side as an independent variable, and the specific formula is as follows:
in the formula 6, s.t. is a constraint condition, and I is the generating income of the generating side of the market at the day before; c is the power generation cost of the power generation side of the market at the day before; r total In order to maximize the generation income of the day-ahead market generation side.
The marginal price of the day-ahead market node of the ith unit in the nth scene at the time t; q i，t Generating capacity of the ith unit at the moment t; a is a i 、b i And c i The cost coefficient of the unit i; q i，t，s The winning electricity amount of the ith unit in the s section at the time t; q t min And Q t max Respectively the minimum/large power generation amount at the power generation side at the time t; q i min And Q i max Respectively the minimum/large power generation amount of the i unit; mu.s t Is the frequency modulation coefficient and has a value of 0 or 1 if mu t If the power generation capacity of the ith unit is 0, the power generation capacity of the ith unit approaches to the minimum output at the moment t; on the contrary, the maximum output is approached; f. of i down And f i up And respectively reserving down/up frequency modulation capacity for the ith unit. Note that: if the unit i does not select frequency modulation, f i down And f i up Are both 0; r i min And R i max The minimum/large climbing amount of the i unit is respectively.
And
reporting and quoting the s section of the ith unit; p is a radical of b，max And p b，min Respectively is the upper value and the lower value of the quoted price;
is the minimum report value for each segment.
The following describes 12 constraint equations in the constraint: the 1 st constraint formula and the 2 nd constraint formula respectively represent the generating income and the generating cost of all the units in one day; the 2 nd formula represents the sum of the bid amount corresponding to the quotation curve; the 4 th constraint formula and the 5 th constraint formula represent that the power generation side is limited by the power generation amount at the time t; the 6 th constraint formula represents that the unit is limited by the power generation capacity; the 7 th constraint formula represents that the unit is limited by climbing; the 8 th constraint formula represents a medium bid amount constraint. The meaning is as follows: if the quoted price of the ith unit at the s section is not less than the node marginal electricity price, the bid amount of the ith unit at the s section at the t moment is correspondingly the report amount of the ith unit at the s section; otherwise, it is 0. The 9 th constraint formula represents that the quotation curve of the power generation side is required to be monotonically non-decreasing along with the output; the 10 th constraint formula represents that the final declaration value of the quotation curve at the power generation side is the maximum capacity; the 11 th constraint formula represents that the quotation curve is limited by quotation; the 12 th constraint equation represents that the quote curve is volume limited.
Further, the process of optimizing the parameters of the offer decision model by using the dynamic firefly algorithm is as follows:
the method comprises the steps of firstly, carrying out population initialization by adopting uniform distribution, then solving a quotation decision model by adopting a momentum firefly algorithm, and then terminating population search according to search conditions to obtain corresponding optimal quotation curves under different price scenes. Compared with the traditional firefly algorithm, the momentum firefly algorithm can avoid the problems of local optimum, low convergence speed and the like by introducing momentum parameters.
Since the optimization effect of the firefly search algorithm depends on the initial position, the population of each group can be generally distributed uniformly to obtain a randomized initial position, and the specific formula is as follows:
in formula 7, X g min And X g max The upper limit and the lower limit of the firefly g are respectively referred to as a sectional report price; x g 0 Is the initialized position of the firefly g; rand () is [0,1 ]]Random number in between. And (3) carrying out sectional report quotation:
and
reporting and quoting for the s section of the ith unit;
the momentum firefly algorithm is characterized in that momentum parameters are introduced, wherein the momentum refers to gathering past historical position memory, and a specific formula is as follows:
in equations 8 and 9, r gl Is x g And x g The distance between the two; gamma is an attraction coefficient; r is 0 Is r of gl An attractive force equal to 0; alpha is a step-size factor, in the classical firefly algorithmThe step factor is typically set between 0 and 2; e is a vector that follows a gaussian distribution. w is a weight factor; v g k The momentum parameter of the g-th firefly at the k iterations; x g k Is the position of the g-th firefly at the k-th iteration. As can be seen from formulas (23) to (24): if the second term of equation (8) is greatly changed during the search (if the position of the g-th firefly is sharply increased), the search is passed through V g k Weighting reduces V g k+1 The position change is reduced, so that the convergence speed is accelerated by inhibiting the oscillation; on the other hand, if the second term of equation (8) is slightly changed (if the position of the g-th firefly is greatly decreased), the second term passes through V g k Weight is increased by V g k+1 The position change is increased, so that the convergence precision is improved by jumping out of a local convergence mode. When the number of iterations reaches the maximum number, the iteration will stop.
And S308, determining the volume benefit and the quotation scheme under different price scenes by using the constructed quotation decision model.
The cost and the quotation scheme are calculated according to different price scenes, and therefore multi-dimensional comparison is achieved.
According to the embodiment of the application, firstly, a lightweight gradient elevator of an artificial intelligence technology is used for predicting node marginal electricity prices, electricity price distribution of different confidence intervals is obtained according to a quantile regression method, then a quotation decision model is constructed by combining unit physical constraints and price scenes, the optimal parameters of the quotation decision model are solved by combining a momentum firefly algorithm of an industrial intelligence technology, the optimization speed is improved by introducing a momentum theory through the momentum firefly algorithm, and finally, different quantity benevolence schemes are established for comparison according to different electricity price scenes, so that effective evaluation basis is provided for reasonable quotation and risk management of a power generation side.
The following are embodiments of the apparatus of the present application that may be used to perform embodiments of the method of the present application. For details which are not disclosed in the embodiments of the apparatus of the present application, reference is made to the embodiments of the method of the present application.
Referring to fig. 4, a schematic structural diagram of an artificial intelligence-based electricity generation-side price quotation device according to an exemplary embodiment of the present application is shown. The artificial intelligence-based power generation side electricity price quotation device can be realized into all or part of the electronic equipment through software, hardware or a combination of the software and the hardware. The device 4 comprises: selection unit 401, training unit 402, prediction unit 403, and determination unit 404.
A selection unit 401 for selecting a relevant feature among the potentially influencing features of the electricity prices;
a training unit 402, configured to train sample data of the relevant features by using an ensemble learning framework based on a decision tree to obtain a feature training model;
the prediction unit 403 is configured to input data to be predicted into the feature training model to obtain a predicted value determined by the marginal price of the market node in the future;
the predicting unit 403 is further configured to process the determined predicted value of the marginal price of the day-ahead market node by using a quantile regression model to obtain an uncertain predicted value of the marginal price of the day-ahead market node; calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node;
a determining unit 404, configured to randomly generate N price scenarios; wherein N is an integer greater than 1, and the probability of each price scene is 1/N;
and constructing a quotation decision model under each price scene, and determining the volume benefit and quotation scheme under different price scenes by using the constructed quotation decision model.
In one or more possible embodiments, the selecting the relevant feature among the plurality of potentially influencing features of the electricity prices includes:
a related feature is selected among a plurality of potentially influencing features of electricity prices using a lasso algorithm.
In one or more possible embodiments, the training sample data of the relevant features to obtain a feature training model includes:
and training the sample data of the relevant characteristics by using a lightweight gradient elevator to obtain a characteristic training model.
In one or more possible embodiments, the randomly generating N price scenarios includes:
and randomly generating N price scenes by adopting a Monte Carlo simulation method.
In one or more possible embodiments, the constructing a price decision model in each price scenario includes:
and calculating parameters under each price scene by using a dynamic firefly algorithm, and constructing a quotation decision model by using the calculated parameters.
In one or more possible embodiments, the price quote data model is generated based on an objective function and constraints, the constraints including: the system comprises one or more of frequency modulation constraint conditions, declaration rule constraint conditions, generating capacity constraint conditions, unit climbing constraint conditions and medium bid amount constraint conditions, wherein the objective function is used for realizing maximization of generating benefits according to generating cost and generating income.
In one or more possible embodiments, the relevant features include: market price, load of the whole province, outgoing power, load of A type unit and load bidding space in the past day.
It should be noted that, when the power generation-side electricity price quotation device based on artificial intelligence provided in the above embodiment executes the power generation-side electricity price quotation method based on artificial intelligence, only the division of the above function modules is taken as an example, and in practical application, the function distribution may be completed by different function modules according to needs, that is, the internal structure of the device is divided into different function modules, so as to complete all or part of the functions described above. In addition, the power generation side electricity price quotation device based on artificial intelligence provided by the embodiment and the power generation side electricity price quotation method based on artificial intelligence have the same concept, the detailed implementation process is shown in the method embodiment, and the detailed description is omitted here.
The above-mentioned serial numbers of the embodiments of the present application are merely for description, and do not represent the advantages and disadvantages of the embodiments.
An embodiment of the present application further provides a computer storage medium, where the computer storage medium may store a plurality of instructions, where the instructions are suitable for being loaded by a processor and executing the method steps in the embodiment shown in fig. 2, and a specific execution process may refer to a specific description of the embodiment shown in fig. 2, which is not described herein again.
Fig. 5 is a schematic structural diagram of an apparatus according to an embodiment of the present disclosure. As shown in fig. 5, the apparatus may be the artificial intelligence-based power generation-side power price quotation apparatus of fig. 1, and the artificial intelligence-based power generation-side power price quotation apparatus 1000 may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, memory 1005, at least one communication bus 1002.
Wherein a communication bus 1002 is used to enable connective communication between these components.
The user interface 1003 may include a Display screen (Display) and a Camera (Camera), and optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.
The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface), among others.
The Memory 1005 may include a Random Access Memory (RAM) or a Read-Only Memory (Read-Only Memory). Optionally, the memory 1005 includes a non-transitory computer-readable medium. The memory 1005 may be used to store an instruction, a program, code, a set of codes, or a set of instructions. The memory 1005 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the various method embodiments described above, and the like; the storage data area may store data and the like referred to in the above respective method embodiments. The memory 1005 may optionally be at least one memory device located remotely from the processor 1001. As shown in fig. 5, the memory 1005, which is a kind of computer storage medium, may include therein an operating system, a network communication module, a user interface module, and an application program.
In the apparatus 1000 shown in fig. 5, the user interface 1003 is mainly used as an interface for providing input for a user, and acquiring data input by the user; and processor 1001 may be configured to invoke an application program stored in memory 1005 to configure an application program interface and to perform the method embodiments shown in fig. 2 or fig. 3 in particular.
The concept of this embodiment is the same as that of the embodiment of the method in fig. 2 or fig. 3, and the technical effects brought by the embodiment are also the same, and the specific process can refer to the description of the embodiment in fig. 2 or fig. 3, which is not described again here.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by a computer program, which may be stored in a computer readable storage medium and executed by a computer to implement the processes of the embodiments of the methods described above. The storage medium can be a magnetic disk, an optical disk, a read-only memory or a random access memory.
The above disclosure is only for the purpose of illustrating the preferred embodiments of the present application and should not be taken as limiting the scope of the present application, so that the present application will be covered by the appended claims.
Claims (10)
1. A power generation side electricity price quotation method based on artificial intelligence is characterized by comprising the following steps:
selecting a relevant feature among a plurality of potentially influencing features of electricity prices;
training the sample data of the relevant features by using an integrated learning framework based on a decision tree to obtain a feature training model;
inputting data to be predicted into the feature training model to obtain a predicted value for determining the marginal price of the market node at the day before;
processing the day-ahead market node marginal price determination predicted value by using a quantile regression model to obtain a day-ahead market node marginal price uncertainty predicted value;
calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node;
randomly generating N price scenes; wherein N is an integer greater than 1, and the probability of each price scene is 1/N;
and constructing a quotation decision model under each price scene, and determining the volume benefit and quotation scheme under different price scenes by using the constructed quotation decision model.
2. The method of claim 1, wherein selecting the relevant feature among the plurality of potentially influencing features of electricity prices comprises:
a related feature is selected among a plurality of potentially influencing features of electricity prices using a lasso algorithm.
3. The method according to claim 1 or 2, wherein training the sample data of the relevant features results in a feature training model, comprising:
and training the sample data of the relevant characteristics by using a lightweight gradient elevator to obtain a characteristic training model.
4. The method of claim 3, wherein the randomly generating N price scenarios comprises:
and randomly generating N price scenes by adopting a Monte Carlo simulation method.
5. The method according to claim 1 or 2, wherein the constructing of the offer decision model in each price scenario comprises:
and calculating parameters under each price scene by using a dynamic firefly algorithm, and constructing a quotation decision model by using the calculated parameters.
6. The method of claim 5, wherein the offer prediction model is generated based on an objective function and constraints, the constraints comprising: the system comprises one or more of frequency modulation constraint conditions, declaration rule constraint conditions, generating capacity constraint conditions, unit climbing constraint conditions and medium bid amount constraint conditions, wherein the objective function is used for realizing maximization of generating benefits according to generating cost and generating income;
wherein, the quotation decision model is expressed according to the following formula:
R total ＝maxI-C
s.t. is a constraint condition, and I is the generating income of the market generating side in the day ahead; c is the power generation cost of the power generation side of the market at the day before; r is total In order to maximize the generation income of the day-ahead market generation side;
marginal price of day-ahead market node of ith unit under nth scene at time tGrid; q i，t Generating capacity of the ith unit at the moment t; a is a i 、b i And c i The cost coefficient of the unit i is obtained; q i，t，s The winning electricity amount of the ith unit in the s section at the time t; q t min And Q t max Respectively is the minimum generating capacity/large generating capacity of the generating side at the moment t; q i min And Q i max Respectively the minimum/large power generation amount of the i unit; mu.s t Is a frequency modulation coefficient and has a value of 0 or 1; mu.s of t If the power generation capacity of the ith unit is 0, the power generation capacity of the ith unit approaches to the minimum output at the moment t; mu.s of t 1, which approaches the maximum output; f. of i down And f i up Respectively reserving down/up frequency modulation capacity for the ith unit; if the unit i does not select frequency modulation, f i down And f i up Are both 0; r i min And R i max Respectively the minimum/large climbing amount of the i unit;
and
reporting and quoting the s section of the ith unit; p is a radical of b，max And p b，min Respectively is the upper value and the lower value of the quoted price;
is the minimum report value for each segment.
7. The method according to claim 1 or 2, wherein the relevant features comprise: market price, provincial load, delivery power, A-type unit load and load bidding space before the historical date.
8. The utility model provides a power generation side price quotation device based on artificial intelligence which characterized in that includes:
a selection unit for selecting a relevant feature among the potential influence features of the electricity prices;
the training unit is used for training the sample data of the relevant features by utilizing an integrated learning framework based on a decision tree to obtain a feature training model;
the prediction unit is used for inputting data to be predicted into the feature training model to obtain a day-ahead market node marginal price determination prediction value;
the prediction unit is also used for processing the determined predicted value of the marginal price of the day-ahead market node by using a quantile regression model to obtain the uncertain predicted value of the marginal price of the day-ahead market node; calculating an interval predicted value through a preset confidence interval and a quantile regression model based on the uncertain predicted value of the marginal price of the day-ahead market node;
a determining unit for randomly generating N price scenes; wherein N is an integer greater than 1, and the probability of each price scene is 1/N;
and constructing a quotation decision model in each price scene, and determining the volume benefit and quotation scheme in different price scenes by using the constructed quotation decision model.
9. A computer storage medium, characterized in that it stores a plurality of instructions adapted to be loaded by a processor and to perform the method steps according to any one of claims 1 to 7.
10. An electronic device, comprising: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to perform the method steps of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110374116.1A CN113159835B (en) | 2021-04-07 | 2021-04-07 | Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110374116.1A CN113159835B (en) | 2021-04-07 | 2021-04-07 | Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113159835A CN113159835A (en) | 2021-07-23 |
| CN113159835B true CN113159835B (en) | 2023-02-28 |
Family
ID=76889093
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110374116.1A Active CN113159835B (en) | 2021-04-07 | 2021-04-07 | Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113159835B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113763184A (en) * | 2021-08-26 | 2021-12-07 | 甘肃同兴智能科技发展有限责任公司 | Carbon asset assessment method |
| CN113706269A (en) * | 2021-09-13 | 2021-11-26 | 华润电力技术研究院有限公司 | Frequency modulation service quotation method and frequency modulation service quotation device |
| CN113919886A (en) * | 2021-11-11 | 2022-01-11 | 重庆邮电大学 | Data characteristic combination pricing method and system based on summer pril value and electronic equipment |
| CN116308441A (en) * | 2023-03-07 | 2023-06-23 | 太原科技大学 | House price prediction method with variable selection capability |
| CN116433277B (en) * | 2023-06-13 | 2023-09-19 | 国网浙江省电力有限公司 | Electricity price conduction processing method and system based on novel electric power system |
| CN116434029B (en) * | 2023-06-15 | 2023-08-18 | 西南石油大学 | Drinking detection method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106228279A (en) * | 2016-09-06 | 2016-12-14 | 北京华电天仁电力控制技术有限公司 | The electricity price optimized calculation method of electricity market Generation Side based on Protect advantage model |
| CN108711890A (en) * | 2018-06-27 | 2018-10-26 | 广东电网有限责任公司 | Ahead market goes out clearing method, system, device and computer readable storage medium |
| CN109102155A (en) * | 2018-07-09 | 2018-12-28 | 中国南方电网有限责任公司 | A kind of ultra-short term deploying node probability forecasting method and system |
| CN110648164A (en) * | 2019-08-15 | 2020-01-03 | 广东电网有限责任公司电力调度控制中心 | Method and device for determining marginal electricity price interval of node |
| CN110930172A (en) * | 2018-09-19 | 2020-03-27 | 中国电力科学研究院有限公司 | Neural network-based cross-country market joint marginal price prediction method and system |
| CN111553750A (en) * | 2020-05-12 | 2020-08-18 | 太原理工大学 | Energy storage bidding strategy method considering power price uncertainty and loss cost |
| CN111740408A (en) * | 2020-06-19 | 2020-10-02 | 中国电建集团青海省电力设计院有限公司 | Photo-thermal power station optimal quotation decision method based on robust random model |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111210144A (en) * | 2020-01-02 | 2020-05-29 | 南方电网科学研究院有限责任公司 | Power generation risk management method and system for electric power spot market |
| CN112001744A (en) * | 2020-07-28 | 2020-11-27 | 安徽电力交易中心有限公司 | Power generator auxiliary quotation system and method based on prospect theory in electric power spot market |
-
2021
- 2021-04-07 CN CN202110374116.1A patent/CN113159835B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106228279A (en) * | 2016-09-06 | 2016-12-14 | 北京华电天仁电力控制技术有限公司 | The electricity price optimized calculation method of electricity market Generation Side based on Protect advantage model |
| CN108711890A (en) * | 2018-06-27 | 2018-10-26 | 广东电网有限责任公司 | Ahead market goes out clearing method, system, device and computer readable storage medium |
| CN109102155A (en) * | 2018-07-09 | 2018-12-28 | 中国南方电网有限责任公司 | A kind of ultra-short term deploying node probability forecasting method and system |
| CN110930172A (en) * | 2018-09-19 | 2020-03-27 | 中国电力科学研究院有限公司 | Neural network-based cross-country market joint marginal price prediction method and system |
| CN110648164A (en) * | 2019-08-15 | 2020-01-03 | 广东电网有限责任公司电力调度控制中心 | Method and device for determining marginal electricity price interval of node |
| CN111553750A (en) * | 2020-05-12 | 2020-08-18 | 太原理工大学 | Energy storage bidding strategy method considering power price uncertainty and loss cost |
| CN111740408A (en) * | 2020-06-19 | 2020-10-02 | 中国电建集团青海省电力设计院有限公司 | Photo-thermal power station optimal quotation decision method based on robust random model |
Non-Patent Citations (1)
| Title |
|---|
| 基于改进萤火虫算法的电力市场多因素博弈均衡研究;周保荣 等;《南方电网技术》;20200531;第14卷(第3期);第54-61页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113159835A (en) | 2021-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113159835B (en) | Power generation side electricity price quotation method and device based on artificial intelligence, storage medium and electronic equipment | |
| Rashedi et al. | Markov game approach for multi‐agent competitive bidding strategies in electricity market | |
| Zaman et al. | Co-evolutionary approach for strategic bidding in competitive electricity markets | |
| CN109472700A (en) | Prediction technique, server and the storage medium of stock price | |
| Zhou et al. | Four‐level robust model for a virtual power plant in energy and reserve markets | |
| CN112785005B (en) | Multi-objective task assistant decision-making method and device, computer equipment and medium | |
| CN106407305A (en) | Data mining system and method | |
| Wu et al. | Strategic bidding in a competitive electricity market: An intelligent method using Multi-Agent Transfer Learning based on reinforcement learning | |
| Aliabadi et al. | The emerging threat of artificial intelligence on competition in liberalized electricity markets: A deep Q-network approach | |
| Afshar et al. | Fuzzy multi-objective optimization of finance-based scheduling for construction projects with uncertainties in cost | |
| CN114202174A (en) | Electricity price risk grade early warning method and device and storage medium | |
| Bigdeli et al. | Bidding strategy in pay-as-bid markets based on supplier-market interaction analysis | |
| Dong et al. | Bidding strategy of thermal power compound differential evolution game under the market mechanism of peak regulation auxiliary service | |
| CN110555742A (en) | Generation method and system for generator agent quotation | |
| Sharma et al. | An intelligent energy bidding strategy based on opposition theory enabled grey wolf optimizer | |
| Liang et al. | Operation-adversarial scenario generation | |
| Wang et al. | Research on the decision-making method of coal order price and coal purchase quantity based on prediction | |
| CN114118608A (en) | Power grid business matching optimization method and device, terminal and storage medium | |
| Cano-Martínez et al. | Dynamic energy prices for residential users based on Deep Learning prediction models of consumption and renewable generation | |
| CN113379533A (en) | Method, device, equipment and storage medium for improving circulating loan quota | |
| Zhang et al. | Negotiation game model for big data transactions | |
| Karimi et al. | Participation of generating companies in transmission investment in electricity markets | |
| Rogers et al. | A genetic algorithm approach to the smart grid tariff design problem | |
| Weishang et al. | A trading optimization model for virtual power plants in day-ahead power market considering uncertainties | |
| Alcántara et al. | Optimal day-ahead offering strategy for large producers based on market price response learning |
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 |