Search

CN-122026505-A - Multi-energy power system day-ahead optimal scheduling method considering electric-carbon cooperation

CN122026505ACN 122026505 ACN122026505 ACN 122026505ACN-122026505-A

Abstract

The invention discloses a day-ahead optimal scheduling method for a multi-energy power system taking electric-carbon cooperation into consideration, and relates to the field of day-ahead optimal scheduling of the power system in power system planning operation. The method comprises the steps of constructing a new mixed model after inserting an LSTM into a multi-head attention layer of a Transformer, respectively establishing a prediction model suitable for wind power generation power, light power generation power and load power by using the new mixed model, defining an evolution virtual net load, calculating the evolution virtual net load according to a double-track system absorption requirement for renewable energy power generation and a prediction result, which are proposed by a full-guarantee purchasing renewable energy power monitoring method, introducing a dynamic electric carbon emission factor to establish an electricity price and carbon price synergistic relationship, designing a new electric carbon synergistic mechanism, establishing a corresponding electric carbon synergistic model, taking the minimum comprehensive operation cost of the system as a target, establishing a multi-energy power system optimization scheduling model considering electric carbon synergy, and solving the model to obtain a scheduling plan.

Inventors

  • XIAO BAI
  • CHANG YULONG
  • JIANG ZHUO

Assignees

  • 东北电力大学

Dates

Publication Date
20260512
Application Date
20251217

Claims (10)

  1. 1. A day-ahead optimal scheduling method for a multi-energy power system considering electric-carbon cooperation is characterized by comprising the following steps: S1, establishing a wind-solar power generation and load power prediction model based on an improved LSTM-converter; The improved wind-light power generation and load power prediction model of the LSTM-converter is coupled with the time sequence memory characteristic of the LSTM and the multi-head attention mechanism of the converter, and the wind-light power generation power and the total load power are predicted through an input link, an encoding link, a decoding link and an output link; Step S2, calculating an evolution virtual payload; calculating an evolution virtual payload by using the data predicted in the step S1, wherein the evolution virtual payload is defined as the evolution virtual payload of a period of time after the power and the minimum technical output of the thermal power corresponding to the guaranteed purchase power in the renewable energy power generation are subtracted from the total load of the period of time in a power system comprising a wind power plant, a photovoltaic power generating unit, a thermal power generating unit and a hydroelectric generating unit; S3, establishing an electric carbon cooperative model; and (3) constructing an electricity-carbon cooperative model of time-of-use electricity price and time-of-use carbon price by utilizing the evolution virtual payload calculated in the step (S2), wherein the method comprises the following steps of: S31, dividing a total scheduling period into 24 price time periods, and calculating a dynamic electricity carbon emission factor D-ECEF in the total scheduling period; Step S32, normalizing the D-ECEF and thermal power total power and the system up-regulation flexibility margin; s33, clustering based on a K-means algorithm, and dividing peak time periods, normal time periods and valley time periods of carbon prices and electricity prices according to clustering results; step S34, obtaining time-sharing electricity price and time-sharing carbon price, and calculating the electricity price and the carbon price of the t scheduling period; step S35, a demand response model and a carbon transaction model are established on the basis of the time-sharing electricity price and the time-sharing carbon price obtained in the step S34; step S4, establishing a day-ahead optimal scheduling model of the multi-energy power system considering electric-carbon cooperation; The scheduling model utilizes the evolution virtual net load of the step S2 and the electric carbon cooperative model of the step S3, and aims at the total system cost C, and an objective function is established through the total operation cost C g of the thermal power unit, the operation cost C h of the hydroelectric power unit, the operation cost C w of the wind power plant, the operation cost C v of the photovoltaic power station and the carbon emission cost C c ; The constraint conditions comprise thermal power unit operation constraint, hydroelectric unit operation constraint, wind power station constraint, photovoltaic power station constraint, power balance constraint, up-regulation flexibility constraint and down-regulation flexibility constraint; Step S5, solving the scheduling model of the step S4 by utilizing the evolution virtual payload of the step S2 and under the condition of fixed carbon price, and calculating D-ECEF according to the obtained scheduling result; Step S6, firstly, generating a time-of-use electricity price and a time-of-use carbon price by combining the supply and demand relation of electricity and carbon by using the D-ECEF calculated in the step S5, secondly, recalculating an evolution virtual net load of the step S2 by taking the time-of-use electricity price as input of a demand response, then solving a scheduling model of the step S4 under the condition of the new evolution virtual net load and the time-of-use carbon price, and finally, calculating a new D-ECEF according to a new scheduling result; Step S7, firstly, generating a new time-sharing electricity price and a new time-sharing carbon price by combining the new D-ECEF calculated in the step S6 with the supply and demand relation of electricity and carbon respectively, secondly, recalculating the evolution virtual net load of the step S2 by taking the time-sharing electricity price as the input of a demand response, and finally solving the scheduling model of the step S4 under the condition of the evolution virtual net load and the time-sharing carbon price and the condition of the new time-sharing carbon price, thereby obtaining the multi-energy power system scheduling plan considering the electricity-carbon cooperation.
  2. 2. The day-ahead optimal scheduling method of the multi-energy power system taking into account the electric-carbon cooperation of claim 1, wherein the input link of the step S1 is to normalize each original power time sequence, and then map the power time sequence to a high-dimensional space through a coding embedding layer, so that the attention mechanism of the model learns the time dependence relationship in the power time sequence; The coding link comprises a multi-head attention layer, an LSTM layer, a residual error connection and normalization layer, a feedforward network layer, a non-linear activation function, a power sequence characterization and a decoder, wherein the multi-head attention layer is firstly used for synchronously capturing different time scale characteristics of a plurality of subspaces in a power time sequence through a plurality of groups of attention mechanisms calculated in parallel, and directly establishing a global dependency relationship between remote time points; The decoding link adopts a double-attention mechanism structure, wherein the first layer is a mask multi-head attention layer, receives the input of a decoder, uses the mask mechanism to prevent the current position from accessing the information of the future time steps, ensures the causality of the autoregressive generating process, and outputs a power time sequence to the second layer multi-head attention mechanism layer through LSTM processing, residual connection and normalization operation; And the output link maps the high-dimensional power time sequence vector output by the decoder to a low-dimensional space through the linear layer, converts the output of the linear layer into probability distribution by using a Softmax activation function, and selects the result with the highest probability as a power predicted value to output.
  3. 3. The day-ahead optimal scheduling method for the multi-energy power system taking into account the electric-carbon coordination according to claim 1, wherein the calculation formula of the evolving virtual payload in the step S2 is as follows: Wherein, P VL,t is the evolution virtual net load of the t scheduling period, P L,t is the total load of the t scheduling period, beta w and beta v are the guaranteed acquisition coefficients of wind power and photovoltaic respectively; And Maximum power of wind power and photovoltaic power generation in the t scheduling period respectively; Beta h is the guaranteed acquisition coefficient of the hydropower, which is the minimum power of the kth thermal power unit in the t scheduling period; and N g and N h are the numbers of thermal power units and hydroelectric units in the system respectively.
  4. 4. The day-ahead optimal scheduling method for the multi-energy power system with consideration of electric-carbon coordination according to claim 1, wherein the calculation formula of the D-ECEF in one total scheduling period in the step S31 is as follows: the method comprises the steps of enabling N i to be an electric carbon emission factor of an ith price period, enabling E r,i to be actual carbon emission of a thermal power unit of the ith price period, enabling P w,i and P v,i to be wind power and photovoltaic power generation power of the ith price period respectively, enabling P g,k,i and P h,k,i to be power generation power of a kth hydroelectric power unit and power generation power of a kth thermal power unit of the ith price period respectively, enabling P j,i to be power which is transmitted to the region in a net mode in the ith price period, and enabling E j,i to be carbon emission intensity of the region j in the ith price period.
  5. 5. The day-ahead optimal scheduling method for the multi-energy power system considering electric-carbon coordination according to claim 1, wherein the formula for normalizing the D-ECEF, the total power of thermal power and the system up-regulation flexibility margin in step S32 is as follows: In the formula, N max and N min are respectively the maximum value and the minimum value of D-ECEF in a total scheduling period, and P g,i is the total power generation amount of the thermal power unit in the ith price period; the normalized value of the total power generation amount of the thermal power generating unit in the ith price period; And The maximum value and the minimum value of the total power generation amount of the thermal power generating unit in a total scheduling period are set; an up-regulation flexibility margin for the ith price period system; And M u,max and M u,min are respectively the maximum value and the minimum value of the up-regulation flexibility margin in one total scheduling period.
  6. 6. The day-ahead optimal scheduling method of the multi-energy power system considering electric-carbon coordination according to claim 1 is characterized in that in the step S33, the K-means algorithm performs clustering division by judging Euclidean distances between different parameters and clustering center points, wherein D-ECEF normalized values and thermal power unit power sum normalized values are used as sample sets for clustering analysis when dividing carbon price time periods, the clustering category number is set to 3, the clustering center points select D-ECEF normalized values as1, the thermal power unit power sum normalized values as1 as peak points, D-ECEF normalized values as 0.5, the thermal power unit power sum normalized values as 0.5 as median points, and D-ECEF normalized values as 0 and thermal power unit power sum normalized values as 0 as valley points; Dividing the electricity price time period, taking the normalized value of the D-ECEF and the normalized value of the up-regulation flexibility margin of each time period as a sample set of cluster analysis, and changing the initial point of a cluster center point into the peak point of (1, 0), the median point of (0.5 ) and the valley point of (0, 1).
  7. 7. The day-ahead optimal scheduling method for the multi-energy power system taking into account the electric-carbon coordination according to claim 1, wherein the calculation formula of the electricity price and the carbon price of the t scheduling period in the step S34 is as follows: In the formula, And The electricity price and the carbon price of the system are respectively the t scheduling period; peak time electricity price, normal time electricity price and valley time electricity price respectively; peak time carbon price, ordinary time carbon price and valley time carbon price respectively; As a state variable, when the t-th scheduling period is at the peak-time electricity price period 1, 0 At other periods; as a state variable, when the t-th scheduling period is in the flat electricity price period 1, 0 At other periods; As a state variable, when the t-th scheduling period is in the off-peak electricity price period 1, 0 At other periods; As a state variable, when the t-th scheduling period is at the peak-time carbon price period 1, 0 At other periods; as a state variable, when the t-th scheduling period is at the usual carbon price period 1, 0 At other periods; As a state variable, when the t-th scheduling period is at the valley time carbon price period 1, 0 At other periods.
  8. 8. The day-ahead optimal scheduling method of the multi-energy power system considering electric-carbon coordination according to claim 1, wherein the demand response model in the step S35 is based on price demand response of time-of-use electricity price, and the sensitivity of the electricity consumption to price in different periods of peak, flat and valley is represented by using an electricity consumption price elastic matrix, where the electricity consumption price elastic matrix is: Wherein epsilon is an electric quantity electricity price elastic matrix, epsilon ff 、ε pp and epsilon gg are respectively the self-elastic coefficients of peak, flat and valley periods, epsilon fp is the elastic coefficient of the electric quantity of the peak period to the electric quantity of the ordinary period, epsilon fg is the elastic coefficient of the electric quantity of the peak period to the electric quantity of the valley period, epsilon pf is the elastic coefficient of the electric quantity of the ordinary period Duan Dianliang to the electric quantity of the peak period, epsilon pg is the elastic coefficient of the electric quantity of the ordinary period to the electric quantity of the valley period, epsilon gf is the elastic coefficient of the electric quantity of the valley period to the electric quantity of the peak period, and epsilon gp is the elastic coefficient of the electric quantity of the valley period to the electric quantity of the ordinary period; after the demand response, the power consumption expressions of the peak, the flat and the valley periods are as follows: Wherein q 0 、q 1 represents the electricity consumption of each period before and after the implementation of the demand response, q 0,f 、q 0,p 、q 0,g represents the electricity consumption of the periods of the peak, the flat and the valley before the implementation of the demand response, and K e represents the fixed electricity price before the implementation of the time-of-use electricity price; The electricity price difference values of the peak, flat and valley periods are respectively represented; the constraint conditions are as follows: Wherein Δp z,t and Δp j,t are the electricity usage increment and decrement amount, respectively, for the t-th scheduling period; And U z,t 、U j,t is a state variable, U z,t is 1 when the power consumption of the t scheduling period is increased, and U j,t is 1 when the power consumption is reduced; In the carbon transaction model, the carbon emission of the system is considered to be all derived from the thermal power unit, the free carbon quota obtained by the thermal power unit is determined by the regional carbon emission permission factor and the power in the period, and the calculation formula is as follows: Wherein E f,t is free carbon quota of the thermal power unit in the T scheduling period, eta is regional carbon emission permission factor, E f is free carbon quota of the thermal power unit in a total scheduling period, P g,k is power sum of all thermal power units in the k scheduling period, and T is total scheduling period; the carbon transaction cost is as follows: Wherein, C c is the system carbon transaction cost, E k is the carbon emission intensity of the kth thermal power generating unit, and E r,t is the actual carbon emission amount of the system in the t scheduling period.
  9. 9. The day-ahead optimal scheduling method for the multi-energy power system considering electric-carbon coordination according to claim 1, wherein the objective function of the scheduling model in step S4 is: minC=C g +C h +C c +C w +C v Wherein, C is the total cost of the system, C g is the total operation cost of the thermal power unit, C h is the operation cost of the hydroelectric unit, C w is the operation cost of the wind farm, and C v is the operation cost of the photovoltaic power station; Wherein, thermal power unit running cost C g is: Wherein, C qt is the total start-stop cost of the thermal power unit, C qt,k is the start-stop cost of the kth thermal power unit, U k,t is the state variable of the kth thermal power unit in the t scheduling period, U k,t =1 represents the unit as an operation state, U k,t =0 represents the unit as a shutdown state, C f is the total coal-burning cost of the thermal power unit, and a k ,b k and C k are the cost coefficients of the kth thermal power unit respectively; The running cost C h of the hydroelectric generating set is as follows: Wherein epsilon h is the unit operation cost of the hydroelectric generating set, P h,k,t is the power generation of the hydroelectric generating set k in the period of t; The wind farm running cost C w is: wherein epsilon w is the running cost of a wind power generation unit, P w,t is the power generation power of the wind power station in the t scheduling period; The photovoltaic power station operation cost C v is: Wherein epsilon v is the running cost of a photovoltaic power generation unit, P v,t is the power generation power of the photovoltaic power station in the t scheduling period; The constraint conditions are specifically as follows: the operation constraint of the thermal power generating unit is as follows: wherein: And The upper limit and the lower limit of the generated power of the kth thermal power generating unit in the t scheduling period are respectively set; And The upper limit and the lower limit of the climbing power of the kth thermal power generating unit are respectively set; And The continuous start-up and stop time of the kth thermal power generating unit at the t-1 moment are respectively; And The minimum startup and shutdown time of the kth thermal power generating unit are respectively; the operation constraint of the hydroelectric generating set is as follows: wherein: the corresponding power of the electric quantity is traded for the kth hydroelectric generating set in the market of the t scheduling period; a lower limit of the generated power of the kth hydroelectric generating set in the t scheduling period; And The upper limit and the lower limit of the climbing power of the kth hydroelectric generating set are respectively; The wind farm constraints are: wherein: trading electric quantity corresponding power for the market of the wind power in the t scheduling period; The photovoltaic power station is constrained as follows: In the formula, The corresponding power of the electric quantity is traded for the photovoltaic market in the t scheduling period; The power balance constraint is: wherein: peak regulation power of the kth thermal power generating unit in the t scheduling period; the flexibility constraint is divided into an up-regulation flexibility constraint and a down-regulation flexibility constraint, which are respectively: wherein, P t su and P t ru are respectively the up-regulation flexibility supply and demand of the system in the t scheduling period; And The method comprises the steps of respectively supplying up-regulating flexibility of a kth thermal power generating unit and a hydroelectric generating unit in a t scheduling period, wherein DeltaP VL,t is the variation of an evolution virtual net load in the t scheduling period and a next scheduling period, and zeta u 、ψ u 、ζ u is the up-regulating flexibility demand coefficient of load, wind power and photovoltaic generated by a prediction error, and zeta e is the up-regulating flexibility demand coefficient of the generating unit during fault maintenance; Wherein, P t sd and P t rd are respectively the down-regulation flexibility supply and demand of the system in the t scheduling period; And And xi d 、ψ d 、ζ d is the load, wind power and photovoltaic down-regulation flexibility demand coefficient generated by the prediction error.
  10. 10. The day-ahead optimal scheduling method of the multi-energy power system with consideration of electric-carbon coordination according to claim 1, wherein a total scheduling period in the step S31 is 24 hours.

Description

Multi-energy power system day-ahead optimal scheduling method considering electric-carbon cooperation Technical Field The invention relates to the field of daily optimization scheduling of a power system in power system planning operation, in particular to a daily optimization scheduling method of a multi-energy power system considering electric-carbon cooperation. Background The existing multi-energy power system scheduling method considering electricity-carbon cooperation has the following problems that although carbon transaction cost is brought into a unit combination decision at a supply side level, a pricing mechanism of the multi-energy power system scheduling method is difficult to reflect the cooperation effect of a carbon transaction market and an electric power market, on a demand side level, although the carbon cost is internalized in the electric power market, time scale cutting, price signal asynchronization and other cooperation barriers still exist between the electric power market and the carbon transaction market, on a system operation level, along with implementation of a full guarantee purchase renewable energy electric quantity supervision method (hereinafter referred to as a method), a renewable energy consumption mechanism is converted from the full guarantee mode of consumption to the double-track purchase combined with market transaction, and the traditional power system scheduling method is not applicable any more. Therefore, the daily optimization scheduling method for the multi-energy power system comprehensively considering wind-solar power generation and load power prediction, evolving virtual net load and electric-carbon cooperation has important practical significance. The invention provides a daily optimization scheduling method of a multi-energy power system taking electric-carbon cooperation into consideration, which constructs a new LSTM-converter neural network, further respectively establishes a prediction model of wind power, light power generation power and load power by using the new LSTM-converter neural network and predicts the wind power, the light power generation power and the load power, defines an evolution virtual payload according to the requirement of a method to generate an evolution virtual payload curve, designs a cooperation mechanism of an electric market and a carbon transaction market, establishes a multi-energy power system optimization scheduling model taking the electric-carbon cooperation into consideration, solves to obtain a scheduling plan, and provides a more reasonable scheduling scheme for daily optimization scheduling of the power system. Disclosure of Invention The invention aims to solve the problems of electric carbon cooperative mechanism deficiency, renewable energy double-track system digestion and renewable energy power consumption in a full-guarantee purchasing mode faced by power system dispatching under a double-carbon target, and provides a multi-energy power system day-ahead optimal dispatching method. The technical scheme adopted by the invention is that the day-ahead optimal scheduling method of the multi-energy power system considering the cooperation of electricity and carbon comprises the following steps: S1, establishing a wind-solar power generation and load power prediction model based on an improved LSTM-converter; The improved wind-light power generation and load power prediction model of the LSTM-converter is coupled with the time sequence memory characteristic of the LSTM and the multi-head attention mechanism of the converter, and the wind-light power generation power and the total load power are predicted through an input link, an encoding link, a decoding link and an output link; Step S2, calculating an evolution virtual payload; calculating an evolution virtual payload by using the data predicted in the step S1, wherein the evolution virtual payload is defined as the evolution virtual payload of a period of time after the power and the minimum technical output of the thermal power corresponding to the guaranteed purchase power in the renewable energy power generation are subtracted from the total load of the period of time in a power system comprising a wind power plant, a photovoltaic power generating unit, a thermal power generating unit and a hydroelectric generating unit; S3, establishing an electric carbon cooperative model; and (3) constructing an electricity-carbon cooperative model of time-of-use electricity price and time-of-use carbon price by utilizing the evolution virtual payload calculated in the step (S2), wherein the method comprises the following steps of: S31, dividing a total scheduling period into 24 price time periods, and calculating a dynamic electricity carbon emission factor D-ECEF in the total scheduling period; Step S32, normalizing the D-ECEF and thermal power total power and the system up-regulation flexibility margin; s33, clustering based on a K-means algorithm, and dividing peak t