CN115622146A - Scheduling decision method for cascade water-light storage complementary system - Google Patents
Scheduling decision method for cascade water-light storage complementary system Download PDFInfo
- Publication number
- CN115622146A CN115622146A CN202211389913.8A CN202211389913A CN115622146A CN 115622146 A CN115622146 A CN 115622146A CN 202211389913 A CN202211389913 A CN 202211389913A CN 115622146 A CN115622146 A CN 115622146A
- Authority
- CN
- China
- Prior art keywords
- storage
- grid
- water
- light
- cascade
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
- H02J3/466—Scheduling the operation of the generators, e.g. connecting or disconnecting generators to meet a given demand
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J15/00—Systems for storing electric energy
- H02J15/003—Systems for storing electric energy in the form of hydraulic energy
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/28—Arrangements for balancing of the load in a network by storage of energy
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a scheduling decision method of a cascade water light storage complementary system, which relates to the technical field of energy comprehensive utilization and comprises the following steps: taking photovoltaic power generation side fluctuation, grid connection point fluctuation and economy as cascade water light storage scheduling optimization targets, and building a cascade water light storage scheduling model by combining grid connection point exchange power constraint, hydropower station and pumped storage reservoir water quantity constraint, node voltage and feeder current constraint; converting the cascade water light and storage scheduling model into a Markov decision process; building a cascade water light storage dynamic scheduling frame based on reinforcement learning; and taking the data of the current cascade water-light-storage complementary system as input, solving a cascade water-light-storage scheduling model converted into a Markov decision process by using a DDPG algorithm, and outputting to obtain a cascade water-light-storage system real-time scheduling strategy corresponding to strong randomness photovoltaic output. The method has high calculation efficiency, greatly relieves the fluctuation rate of the power supply side, improves the system delivery scheduling capability, realizes photovoltaic full-allowance consumption, and can realize real-time decision according to the water-light random environment.
Description
Technical Field
The invention relates to the technical field of comprehensive utilization of energy, in particular to a scheduling decision method of a cascade water-light storage complementary system.
Background
The extreme randomness of photovoltaic output, the uncertainty of water and electricity incoming water and the strong coupling of cascade water and electricity output provide a serious challenge for the operation of a cascade water and light storage system. At present, research on multi-energy complementary cooperative power generation technology at home and abroad has achieved some achievements, aiming at the randomness characteristic of output of renewable energy, conventional power supplies with flexible adjusting capacity, such as hydroelectric power, thermal power and the like or energy storage equipment, are utilized, renewable energy is complemented through advanced adjusting and controlling means, the utilization rate of renewable energy is improved, multi-energy complementary combined power generation is realized, and reference is provided for operation of a cascade water-light storage system. However, most of the existing researches are focused on the complementation of a single hydropower station with large capacity and photovoltaic, and with the cascade development of a river basin for many years, a plurality of cascade small hydropower stations are gradually planned and built, and because the small hydropower stations have limited storage capacity and radial flow type hydropower stations have no regulating capacity, the existing traditional water-light complementary system power generation technology cannot be applied to the cascade water-light storage system. In addition, the traditional optimized scheduling mainly focuses on long-term or day-ahead scheduling, and has no great significance for real-time scheduling decisions, and the related scheduling decision methods are difficult to accurately adapt to the random dynamic changes of the photovoltaic and incoming water, so that the problem that the real-time scheduling decisions of the cascade water-light storage system are urgently needed to be solved by considering the extremely strong randomness of water light.
Disclosure of Invention
The invention provides a scheduling decision method of a cascade water light storage complementary system, which can alleviate the problems.
In order to alleviate the above problems, the technical scheme adopted by the invention is as follows:
the invention provides a scheduling decision method of a cascade water light storage complementary system, which comprises the following steps:
s1, taking photovoltaic power generation side fluctuation, grid-connected point fluctuation and economy as cascade water-light storage scheduling optimization targets, and building a cascade water-light storage scheduling model by combining grid-connected point exchange power constraint, hydropower station and pumped storage reservoir water quantity constraint, node voltage and feeder current constraint;
s2, converting the cascade water light and storage dispatching model into a Markov decision process, and building a cascade water light and storage dynamic dispatching frame based on reinforcement learning;
and S3, under a cascade water light storage dynamic scheduling framework based on reinforcement learning, taking the data of the current cascade water light storage complementary system as input, solving a cascade water light storage scheduling model converted into a Markov decision process by utilizing a Deep Deterministic Policy Gradient (DDPG) algorithm, and outputting to obtain a cascade water light storage system real-time scheduling strategy corresponding to strong stochastic photovoltaic output.
In a preferred embodiment of the present invention, in step S1, the method for optimizing the cascade water light storage dispatching includes dividing the calculation cycle into M stages, and constructing the objective function of optimizing the cascade water light storage dispatching according to the maximization of economic benefit, the minimization of fluctuation on the photovoltaic power generation source side, and the minimization of fluctuation on the grid-connected point
Where F is the total target, ER, in the calculation period T t For the economic benefit, delta P, of the step water light storage system at the time t source,t For source side fluctuation measure at time t, Δ P t Is a grid-connected point fluctuation metric value, beta, in a delta t period 1 、β 2 、β 3 The weighting factors of the economic target, the photovoltaic power generation source side fluctuation stabilizing target and the grid-connected point fluctuation stabilizing target are respectively.
In a preferred embodiment of the invention, the weight factors of the economic target, the photovoltaic power generation source side fluctuation stabilizing target and the grid-connected point fluctuation stabilizing target are calculated by adopting an information entropy theory.
In a preferred embodiment of the invention, the objective of optimizing the photovoltaic generator side fluctuations is to minimize the photovoltaic output fluctuations at each stage of the calculation cycle, calculating the fluctuation measure Δ P at time t source,t The calculation formula of (2) is as follows:
wherein, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The generated output at the moment t of the ith hydropower station, N is the number of the hydropower stations,
the average value of the water luminous output set in the r stage of the period is calculated.
In a preferred embodiment of the present invention, the optimization goal of the grid-connected point fluctuation is to minimize the grid-connected point power fluctuation, form a schedulable outgoing curve, and measure the grid-connected point fluctuation value Δ P within the Δ t period t The calculation formula of (2) is as follows:
ΔP t ＝(P grid,t -P′ grid,t -(P grid,t-1 -P′ grid,t-1 )) 2 ，
wherein, P grid,t The interaction power between the inner ladder level water light storage system and the outer net at the moment t, P hydro,i,t The generated output at the t moment of the ith hydropower station, N is the number of the hydropower stations, P PHS,t For the pumped storage output, P, during time t grid,t The interaction power between the inner ladder level water light storage system and the outer net at the moment t, P grid,t-1 The interaction power between the inner ladder level water light storage system and the outer net at the moment of t-1, P load,t Is load demand at time t, P' grid,t For the pumped storage to participate in the regulation of the interactive power, P ', of the front grid-connected point at time t' grid,t-1 For the time t-1, the interactive power, delta P, of the grid-connected point before the participation of the pumped storage is regulated t The fluctuation metric of the grid-connected point is the delta t period.
In a preferred embodiment of the present invention, the optimization goal of the economy is to make the step water light storage system trade with the external network to obtain the maximum economic benefit, and in the real-time electricity price mode, the step water light storage system is used at the time of tEconomic profit ER t The calculation formula of (2) is as follows:
wherein λ is t For the time t of electricity prices, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The generated output at the t moment of the ith hydropower station, N is the number of the hydropower stations, P PHS,t For the pumped-storage force, P, during time t load,t Load demand at time t.
In a preferred embodiment of the present invention, in step S1, the exchange power constraint of the point-of-connection is
P grid,min ≤P grid,t ≤P grid,max ，
Wherein, P grid,min ，P grid,max Respectively representing the minimum and maximum values of the transmission power of the grid-connected point.
In a preferred embodiment of the invention, the hydropower station and pumped storage reservoir water volume are constrained to
SOC hydro,i,t ＝V i,t /V i,max ，
SOC PHS,t ＝V PHS,t /V PHS,max ，
SOC hydro,i,min ≤SOC hydro,i,t ≤SOC hydro,i,max ，
SOC PHS,min ≤SOC PHS,t ≤SOC PHS,max ，
Wherein, V i,t 、V PHS,t The reservoir capacity, V, of the cascade hydropower station and the pumped storage at time t i,max ，V PHS,max Is the maximum value of the reservoir and the pumped storage water storage capacity of the i-step hydropower station, and is the SOC hydro,i,t 、SOC PHS,t The state of charge, SOC, of the reservoir water of the ith cascade hydropower station and the pumped storage power station respectively hydro,i,max 、SOC hydro,i,min Is the maximum value and the minimum value of the water capacity and the state of charge of the reservoir of the ith cascade hydropower station, and the SOC PHS,max And SOC PHS,min Respectively the maximum value and the minimum value of the water quantity charge state of the reservoir of the pumped storage power station.
In a preferred embodiment of the invention, the node voltage and the feeder current are constrained to
U i,min ≤U i,t ≤U i,max ，
I j,min ≤I j,t ≤I j,max ，
In the formula of U i,t Is the voltage of the I-node at time t, I j,t For the current of the j-th feeder line at time t, V i,min 、V i,max Respectively, the minimum value and the maximum value allowed by the voltage of the I node, I j,min 、I j,max The allowable minimum value and the allowable maximum value of the jth feeder current are respectively.
In a preferred embodiment of the present invention, in step S3, the current step water and light storage complementary system data includes photovoltaic output data, load demand data, electricity price data, and step water and electricity incoming water data; the data of the current cascade water-light-storage complementary system is divided into a training data set and a testing data set, a cascade water-light-storage scheduling model in a Markov decision process is trained and converted by using the training data set, converged model network parameters are stored, and a scheduling decision result of the testing data is obtained by using a converged model network, namely a cascade water-light-storage system real-time scheduling strategy for strong-randomness photovoltaic output is responded.
Compared with the prior art, the invention has the beneficial effects that:
a power supply side staged fluctuation control strategy is provided, so that the inaccuracy of a final scheduling strategy caused by the output difference of a photovoltaic light-rich area and a photovoltaic light-deficient area is avoided; in consideration of the delivery schedulability of the cascade water light storage system, a cascade water light storage system scheduling model capable of fully absorbing the photovoltaic is constructed; designing a real-time interaction environment of Deep reinforcement learning and a cascade water light storage system scheduling model, and solving by using a Deep Deterministic Policy Gradient (DDPG) algorithm to obtain a dynamic scheduling strategy capable of coping with source load random fluctuation changes; from the application perspective, the method greatly relieves the power supply side fluctuation rate, improves the system delivery scheduling capability, meets the grid-connected point power fluctuation rate index requirement, realizes photovoltaic full-allowance consumption, has high calculation efficiency, and can realize real-time decision according to the water-light random environment.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of a scheduling decision method of a cascade water light storage complementary system;
FIG. 2 is a flow chart of a dynamic scheduling solution of a DDPG-based cascade water light storage system;
FIG. 3 is a comparison graph of the average fluctuation on the power supply side;
FIG. 4 is a comparison graph of grid-connected point fluctuations before and after the pumped storage is involved in regulation.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the invention provides a scheduling decision method for a cascade water-light-storage complementary system, which comprises the following steps:
s1, building a cascade water light storage dispatching model, wherein the cascade water light storage dispatching model comprises a cascade water light storage dispatching optimization objective function and constraint conditions.
The cascade water light storage scheduling model is constructed, hydropower is adopted as a complementary power supply aiming at the random fluctuation of photovoltaic power generation, photovoltaic fluctuation is stabilized, and the friendliness of photovoltaic access to a power grid is improved. Because the small hydropower station has limited regulating capacity, the fluctuation of the connecting line is stabilized through pumping storage and cascade hydropower coordination control, the delivery schedulability is improved, and the economic benefit of the whole cascade water light storage system is considered.
1. Optimization objective function for cascade water light storage dispatching
(1) Power supply side ripple stabilization
Considering the existence of rich areas and poor areas of the photovoltaic, the fluctuation of the source side covers the fluctuation caused by randomness and intermittence, the fluctuation is stabilized in the future, the complementary target of the water light is generally set to be a straight line, the average output of the water light period can be taken as a reference to be used as the measurement of the fluctuation, and the fluctuation is shown as the following formula:
wherein, P PV,t Output of photovoltaic power generation at time t, P av The average value of the output of the water photovoltaic power generation in the period is calculated.
However, the purpose of stabilizing the fluctuation is to make the power source have the schedulability acceptable to the power grid, and a straight line is set to stabilize the source-side output, but there are the following problems: (1) The photovoltaic output is extremely random, the output change is large in the day and at night, and the linear target certainly increases the demand of water and electricity capacity; (2) In the method, the small cascade hydropower station has limited regulation capacity, a linear target inevitably causes water abandoning at the midday moment with the maximum photovoltaic output power, and the moment with the small photovoltaic output power cannot meet the photovoltaic fluctuation stabilizing requirement.
In view of the problems, the invention fully considers the existence of rich area and deficient area in photovoltaic, and provides the divisionThe stage fluctuation control strategy is characterized in that a calculation period is divided into M stages, each stage minimizes photovoltaic output fluctuation, the fluctuation measurement of each stage is as the formula, a power supply side stage fluctuation control strategy is provided, the calculation period is divided into M stages, each stage minimizes photovoltaic output fluctuation, and the fluctuation measurement delta P at the t moment in the calculation period source,t The calculation formula of (2) is as follows:
wherein, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The generated output at the moment t of the ith hydropower station is N, which represents the total number of the hydropower stations,
the average value of the water luminous output set in the r stage of the period is calculated.
(2) Grid connection point fluctuation stabilization
In order to improve the schedulability of an outgoing curve and realize source-load matching, the minimum power fluctuation of a grid-connected point is considered as a target to form the schedulable outgoing curve, and a grid-connected point fluctuation metric value delta P in a delta t period t The calculation formula of (2) is as follows:
ΔP t ＝(P grid,t -P′ grid,t -(P grid,t-1 -P′ grid,t-1 )) 2 ，
wherein, P grid,t The interaction power between the inner ladder level water light storage system and the outer net at the moment t, P hydro,i,t The generated output at the t moment of the ith hydropower station, N is the number of the hydropower stations, P PHS,t For the pumped-storage force, P, during time t grid,t For the step water light storage system in the time tInteraction power of system and external network, P grid,t-1 The interaction power between the inner ladder level water light storage system and the outer net at the moment of t-1, P load,t Is load demand at time t, P' grid,t For the pumped storage to participate in the regulation of the interactive power, P ', of the front grid-connected point at time t' grid,t-1 For the time t-1, the interactive power, delta P, of the grid-connected point before the participation of the pumped storage is regulated t And the fluctuation metric value of the grid-connected point is the delta t time interval.
(3) Economy of use
The optimization goal of the economy is to ensure that the cascade water light storage system and the external network trade obtain the maximum economic benefit, and the economic benefit ER of the cascade water light storage system at the time t is realized in a real-time electricity price mode t The calculation formula of (2) is as follows:
wherein, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The generated output at the t moment of the ith hydropower station, N is the number of the hydropower stations, P PHS,t For the pumped-storage force, P, during time t load,t Load demand at time t.
When the cascade water-light storage dispatching optimization objective function is constructed, the stable power output by a cascade water-light storage complementary system is fully considered to ensure friendly access to a power grid, and on the basis, how to obtain the maximum benefit under different runoff conditions is considered, so the cascade water-light storage dispatching optimization objective function is constructed according to the economic benefit maximization, the photovoltaic power generation source side fluctuation minimization and the grid connection point fluctuation minimization
Where F is the total target, ER, in the calculation period T t For the economic benefit, delta P, of the cascade water light storage system at the time t source,t For the value of the fluctuation measure at time t, Δ P t Is a grid-connected point fluctuation metric value, beta, in a delta t period 1 、β 2 、β 3 Respectively an economic target and a photovoltaic power generation source side fluctuation stabilizing targetAnd stabilizing the weight factor of the target by the fluctuation of the mark and grid connection points.
In the invention, the weight factors of an economic target, a photovoltaic power generation source side fluctuation stabilizing target and a grid-connected point fluctuation stabilizing target are calculated by adopting an information entropy theory.
2. Constraint conditions
The cascade water light storage scheduling model needs to meet the exchange power constraint of a grid-connected point, the water quantity constraint of a hydropower station and a pumped storage reservoir, and the node voltage constraint and the feeder current constraint.
(1) And (3) carrying out exchange power constraint of a grid-connected point:
P grid,min ≤P grid,t ≤P grid,max ，
wherein, P grid,min ，P grid,max Respectively representing the minimum and maximum values of the transmission power of the grid-connected point. Exchange power P of grid-connected point grid,t Limited by the transmission capability of the tie line, it is not allowed to exceed its limit.
(2) And (3) restricting the water amount of the hydropower station and the pumped storage reservoir:
SOC hydro,i,t ＝V i,t /V i,max ，
SOC PHS,t ＝V PHS,t /V PHS,max ，
SOC hydro,i,min ≤SOC hydro,i,t ≤SOC hydro,i,max ，
SOC PHS,min ≤SOC PHS,t ≤SOC PHS,max ，
wherein, V i,t 、V PHS,t The reservoir capacity, V, of the cascade hydropower station and the pumped storage at time t i,max ，V PHS,max Is the maximum value of the reservoir and the pumped storage water storage capacity of the i-step hydropower station, and is the SOC hydro,i,t 、SOC PHS,t The state of charge, SOC, of the reservoir water of the ith cascade hydropower station and the pumped storage power station respectively hydro,i,max 、SOC hydro,i,min Is the maximum value and the minimum value of the water capacity and the state of charge of the reservoir of the ith cascade hydropower station, and the SOC PHS,max And SOC PHS,min Respectively the maximum value and the minimum value of the water quantity charge state of the reservoir of the pumping power station.
(3) Node voltage and feeder current constraints:
U i,min ≤U i,t ≤U i,max ，
I j,min ≤I j,t ≤I j,max ，
in the formula of U i,t Is the voltage of the I-node at time t, I j,t For the current of the j-th feeder line at time t, V i,min 、V i,max Respectively, the minimum value and the maximum value allowed by the voltage of the I node, I j,min 、I j,max The allowable minimum value and the allowable maximum value of the jth feeder current are respectively.
And S2, converting the cascade water light and storage dispatching model into a Markov decision process, and building a cascade water light and storage dynamic dispatching framework based on reinforcement learning. The method comprises the steps of constructing actions, states and rewards in the reinforcement learning task conversion process of the cascade water light storage real-time scheduling system.
1. Movement of
The control center of the cascade water light storage complementary dynamic scheduling system is equivalent to an MDP intelligent agent, and the intelligent agent performs real-time state information according to the observed system environment, such as: the method comprises the steps of guiding the cascade water and light storage system to dispatch and operate according to the requirements of economy, source side and grid-connected point power fluctuation stabilization and the like, and enabling cascade water and electricity to generate power p hydro,i,t And pumping and storing power generation/utilization P PHS,t As agent action a t 。
a t ＝{p hydro,i,t ,P PHS,t }，
p hydro,i,t ∈[p hydro,i,min ,p hydro,i,max ]，
P PHS,t ∈[P PHS,min ,P PHS,miax ]，
In the formula, p hydro,i,min 、p hydro,i,max Respectively, the minimum value and the maximum value of the i-level hydroelectric power output, P PHS,min 、P PHS,miax Respectively a minimum and a maximum of the pumped-out force. In the reinforcement learning process, the boundaries of hydroelectric power output and pumped storage output power are limited in the action space, and the constraint condition description is not used.
2. Status of state
Enabling agents to receive corresponding rewards, s, through interaction with the environment t The reinforcement learning main body decides hydropower and pumped storage output power through observed state information for real-time state observation information obtained by continuously interacting with the environment. The states of the cascade water light storage complementary dynamic scheduling system including the time period, the electricity price, the photovoltaic output of the current time period, the load demand, the cascade hydropower station and the pumped storage water capacity State of Charge (SOC) can be described by the following formula:
s t ＝(t,λ t ,P PV,t ,P load,t ,SOC hydro,i,t ,SOC PHS,t )。
3. reward
The cascade water-light storage dynamic scheduling model comprehensively considers the maximization of income and the minimization of source side and grid connection point fluctuation, and obtains the maximum accumulated return through trial and error learning of the control center. However, the optimal strategy must satisfy the constraint conditions in the scheduling model, so that the constraint conditions need to be reasonably converted into partial rewards, which is equivalent to converting an optimization problem with constraint conditions into an optimization problem without constraint conditions, and the reward function is expressed as follows:
r total,t ＝β 1 ER t -β 2 ΔP source,t -β 3 ΔP t ，
wherein, delta 1 Penalty factor, delta, for node voltages exceeding a limit 2 For the penalty factor of branch current exceeding the allowable range, in the present invention, delta is used 1 、δ 2 Set to a constant. The water quantity of reservoir of cascade hydroelectric station or pumped storage power station is similar to the state of charge SOC, delta of battery k,t Corresponding punishment items when the kth SOC exceeds the upper and lower limit rangesThe method comprises the steps of generating a water storage capacity SOC of a hydropower station and a pumped storage, wherein omega is a punishment coefficient.
S3, under a cascade water and light storage dynamic scheduling framework based on reinforcement learning, taking current cascade water and light storage complementary system data (photovoltaic output data, load demand data, electricity price data and cascade water and electricity incoming water data) as input, dividing the current cascade water and light storage complementary system data into a training data set and a testing data set, solving a cascade water and light storage scheduling model converted into a Markov decision process by using a DDPG algorithm (namely, continuously trying and error by using the DDPG algorithm through multiple processes and searching for a scheduling strategy of approaching optimization), and outputting a cascade water and light storage system real-time scheduling strategy corresponding to strong stochastic photovoltaic output, wherein as shown in FIG. 2, the specific flow is as follows:
1) Initializing cascade water light storage scheduling model parameters and DDPG network hyper-parameters, weights and biases.
2) And randomly reading photovoltaic output data, load demand data, electricity price data and cascade hydropower incoming water data of one day in the training data set, and updating the environment model to obtain an initial state.
3) And obtaining actions, namely cascade hydroelectric power and pumped storage output power, based on the current strategy, calculating instant rewards according to the constructed dynamic scheduling environment model of the cascade hydroelectric power storage, and outputting the state of the next moment.
4) A tuple is generated (comprising: status, action, reward, next time status) is stored in the experience pool, the experience pool counter is incremented by one.
5) Judging whether the experience pool is full of memory, if so, selecting L tuples to update the strategy and the value network parameters, then continuing to execute the step 6), and if not, skipping to the step 2);
6) Judging whether the current training is finished, namely if ep is larger than N, judging that all processes are finished and skipping to the step 7), and if not, skipping to the step 2);
7) Outputting accumulated rewards of a plurality of rounds of training, observing whether convergence occurs, if convergence occurs, saving network parameters, and otherwise, skipping to the step 1);
8) And obtaining a scheduling result of the test data by using the stored model network, and outputting the scheduling result.
In the invention, the test data is scheduled and decided by using the convergent model parameters in the training process, and the result after decision can be compared with other traditional methods, such as particle swarm optimization and random planning.
The scheduling decision method of the cascade water-light-storage complementary system is applied in reality, test result analysis is carried out for 6 days optionally, results of source side fluctuation rate and grid-connected point fluctuation rate before and after optimization are shown in figures 3 and 4, although fluctuation rates in different periods are different, after the cascade water power output is dynamically adjusted, the power fluctuation condition of the source side is greatly relieved, the average fluctuation rate per day is reduced from 32.71% to 5.97% and is reduced by about 27%. After the intelligent agent dynamically adjusts the pumping operation condition, the power fluctuation of the grid-connected point is improved, the DDPG intelligent agent can dynamically schedule the pumping output in real time according to the current electricity price, load and the fluctuation condition of the grid-connected point, the average fluctuation rate is reduced from 9.03% to 6.57%, the average fluctuation rate is reduced by about 2.46%, the fluctuation rate index requirement of less than 8% is met, and the photovoltaic full-allowance consumption is realized. DDPG off-line training is time-consuming, but on-line testing can be carried out according to on-line decision of a trained model, and response can reach a second level. The solving time of the example test is shown in table 1, and it can be seen that the solving time of the stochastic programming and the particle swarm algorithm is 16.23s and 90.88s respectively, while the DDPG only needs 0.62s, and the scheduling method in the invention can achieve a second-level decision.
TABLE 1 solving time for different methods
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A scheduling decision method of a cascade water-light storage complementary system is characterized by comprising the following steps:
s1, taking photovoltaic power generation side fluctuation, grid-connected point fluctuation and economy as cascade water light storage scheduling optimization targets, and building a cascade water light storage scheduling model by combining grid-connected point exchange power constraint, hydropower station and pumped storage reservoir water quantity constraint, node voltage and feeder current constraint;
s2, converting the cascade water light and storage dispatching model into a Markov decision process, and building a cascade water light and storage dynamic dispatching frame based on reinforcement learning;
and S3, under a cascade water light storage dynamic scheduling framework based on reinforcement learning, taking the data of the current cascade water light storage complementary system as input, solving a cascade water light storage scheduling model converted into a Markov decision process by utilizing a Deep Deterministic Policy Gradient (DDPG) algorithm, and outputting to obtain a cascade water light storage system real-time scheduling strategy corresponding to strong stochastic photovoltaic output.
2. The scheduling decision method of the cascade water-light-storage complementary system according to claim 1, wherein in step S1, the cascade water-light-storage scheduling optimization method comprises the step of constructing a cascade water-light-storage scheduling optimization objective function according to economic benefit maximization, photovoltaic power generation source side fluctuation minimization and grid connection point fluctuation minimization
Where F is the total target, ER, in the calculation period T t For the economic benefit, delta P, of the step water light storage system at the time t source,t For source side fluctuation measure at time t, Δ P t Is a grid-connected point fluctuation metric value, beta, in a delta t period 1 、β 2 、β 3 The weighting factors of the economic target, the photovoltaic power generation source side fluctuation stabilizing target and the grid-connected point fluctuation stabilizing target are respectively.
3. The scheduling decision method of the cascade water-light-storage complementary system according to claim 2, characterized in that weight factors of an economic objective, a photovoltaic power generation source side fluctuation suppression objective and a grid connection point fluctuation suppression objective are calculated by using an information entropy theory.
4. The dispatch decision method for a cascade water-light-storage complementary system as claimed in claim 2, wherein the objective of optimization of photovoltaic power generation side fluctuation is to minimize photovoltaic output fluctuation at each stage in a calculation cycle, and a fluctuation measure Δ P at time t in the calculation cycle source,t The calculation formula of (2) is as follows:
wherein, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The generated output at the moment t of the ith hydropower station is N, which represents the total number of the hydropower stations,
the average value of the water luminous output set in the r stage of the period is calculated.
5. The scheduling decision method of the cascade water-light-storage complementary system according to claim 4, wherein the optimization goal of the grid-connected point fluctuation is to minimize the grid-connected point power fluctuation, form a schedulable delivery curve, and the grid-connected point fluctuation metric value Δ P in the Δ t period t The calculation formula of (2) is as follows:
ΔP t ＝(P grid,t -P′ grid,t -(P grid,t-1 -P′ grid,t-1 )) 2 ，
wherein, P grid,t The interaction power between the inner ladder level water light storage system and the outer net at the moment t, P hydro,i,t The generated output at the t moment of the ith hydropower station, N is the number of the hydropower stations, P PHS,t For the pumped-storage force, P, during time t grid,t The interaction power between the inner ladder level water light storage system and the outer net at the moment t, P grid,t-1 The interaction power between the inner ladder level water light storage system and the outer net at the moment of t-1, P load,t Is load demand at time t, P' grid,t For the pumped storage to participate in the regulation of the interactive power, P ', of the front grid-connected point at time t' grid,t-1 For the time t-1, the interactive power, delta P, of the grid-connected point before the participation of the pumped storage is regulated t The fluctuation metric of the grid-connected point is the delta t period.
6. The scheduling decision method of the cascade water-light-storage complementary system according to claim 5, wherein the optimization goal of economy is to make the cascade water-light-storage system and the external network trade obtain the maximum economic benefit, and in the real-time electricity price mode, the economic benefit ER of the cascade water-light-storage system at the time t is t The calculation formula of (2) is as follows:
wherein λ is t For electricity price at time t, P PV,t Output of photovoltaic power generation at time t, P hydro,i,t The power generation output at the t moment of the ith hydropower station, N is the number of the hydropower stations, and P is PHS,t For the pumped-storage force, P, during time t load,t Load demand at time t.
7. The scheduling decision method of the step water light storage complementary system according to claim 6, wherein in step S1, the power of the point-of-connection switching is constrained to
P grid,min ≤P grid,t ≤P grid,max ，
Wherein, P grid,min ，P grid,max Respectively representing the minimum and maximum values of the transmission power of the grid-connected point.
8. The method for scheduling decision making of a cascade water light storage complementary system according to claim 7, wherein the water power station and pumped storage reservoir water volume are constrained to
SOC hydro,i,t ＝V i,t /V i,max ，
SOC PHS,t ＝V PHS,t /V PHS,max ，
SOC hydro,i,min ≤SOC hydro,i,t ≤SOC hydro,i,max ，
SOC PHS,min ≤SOC PHS,t ≤SOC PHS,max ，
Wherein, V i,t 、V PHS,t The reservoir capacity, V, of the cascade hydropower station and the pumped storage at time t i,max ，V PHS,max Is the maximum value of the reservoir and the pumped storage water storage capacity of the i-step hydropower station, and is the SOC hydro,i,t 、SOC PHS,t The state of charge, SOC, of the reservoir water of the ith cascade hydropower station and the pumped storage power station respectively hydro,i,max 、SOC hydro,i,min Is the maximum value and the minimum value of the water capacity and the state of charge of the reservoir of the ith cascade hydropower station, and the SOC PHS,max And SOC PHS,min Respectively the maximum value and the minimum value of the water quantity charge state of the reservoir of the pumped storage power station.
9. The step water light storage complementary system scheduling decision method of claim 8, wherein node voltage and feeder current constraints are
U i,min ≤U i,t ≤U i,max ，
I j,min ≤I j,t ≤I j,max ，
In the formula of U i,t Is the voltage of the I-node at time t, I j,t For the current of the j-th feeder line at time t, V i,min 、V i,max Respectively, the minimum value and the maximum value allowed by the voltage of the I node, I j,min 、I j,max The allowable minimum value and the allowable maximum value of the jth feeder current are respectively.
10. The scheduling decision method of the cascade water light storage complementary system according to claim 9, wherein in step S3, the current cascade water light storage complementary system data includes photovoltaic output data, load demand data, electricity price data and cascade water electricity incoming water data; the data of the current cascade water-light-storage complementary system is divided into a training data set and a testing data set, a cascade water-light-storage scheduling model in the Markov decision process is trained and converted by using the training data set, converged model network parameters are stored, and a scheduling decision result of the testing data is obtained by using a converged model network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389913.8A CN115622146A (en) | 2022-11-08 | 2022-11-08 | Scheduling decision method for cascade water-light storage complementary system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389913.8A CN115622146A (en) | 2022-11-08 | 2022-11-08 | Scheduling decision method for cascade water-light storage complementary system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115622146A true CN115622146A (en) | 2023-01-17 |
Family
ID=84878007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211389913.8A Pending CN115622146A (en) | 2022-11-08 | 2022-11-08 | Scheduling decision method for cascade water-light storage complementary system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115622146A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117639111A (en) * | 2024-01-25 | 2024-03-01 | 南京南瑞水利水电科技有限公司 | Photovoltaic fluctuation smooth control method and system based on step radial flow type hydropower |
| CN117674266A (en) * | 2024-01-31 | 2024-03-08 | 国电南瑞科技股份有限公司 | Advanced prediction control method and system for cascade hydropower and photovoltaic cooperative operation |
-
2022
- 2022-11-08 CN CN202211389913.8A patent/CN115622146A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117639111A (en) * | 2024-01-25 | 2024-03-01 | 南京南瑞水利水电科技有限公司 | Photovoltaic fluctuation smooth control method and system based on step radial flow type hydropower |
| CN117639111B (en) * | 2024-01-25 | 2024-04-09 | 南京南瑞水利水电科技有限公司 | Photovoltaic fluctuation smooth control method and system based on step radial flow type hydropower |
| CN117674266A (en) * | 2024-01-31 | 2024-03-08 | 国电南瑞科技股份有限公司 | Advanced prediction control method and system for cascade hydropower and photovoltaic cooperative operation |
| CN117674266B (en) * | 2024-01-31 | 2024-04-26 | 国电南瑞科技股份有限公司 | Advanced prediction control method and system for cascade hydropower and photovoltaic cooperative operation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hossain et al. | Modified PSO algorithm for real-time energy management in grid-connected microgrids | |
| CN107958300B (en) | Multi-microgrid interconnection operation coordination scheduling optimization method considering interactive response | |
| CN107301472B (en) | Distributed photovoltaic planning method based on scene analysis method and voltage regulation strategy | |
| CN106981888B (en) | Wind based on multi-source complementation stores the multiple target dynamic dispatching method of Thermal and Hydroelectric Power Systems | |
| CN112465181A (en) | Two-stage optimization scheduling method supporting source-network-load-storage multi-element ubiquitous coordination | |
| CN115622146A (en) | Scheduling decision method for cascade water-light storage complementary system | |
| CN114336702B (en) | Wind-solar storage station group power distribution collaborative optimization method based on double-layer random programming | |
| CN112800658A (en) | Active power distribution network scheduling method considering source storage load interaction | |
| CN108009693A (en) | Grid-connected micro-capacitance sensor dual blank-holder based on two-stage demand response | |
| CN109034587B (en) | Active power distribution system optimal scheduling method for coordinating multiple controllable units | |
| CN112215433B (en) | Virtual power plant day-ahead optimal scheduling method based on uncertainty of market electricity price | |
| Bonthu et al. | Minimization of building energy cost by optimally managing PV and battery energy storage systems | |
| CN110046780B (en) | Islanding micro-grid demand response economic dispatching method and system based on time-of-use electricity price | |
| CN113765154B (en) | High-elasticity power grid source network load storage multielement collaborative optimization control method | |
| CN115293457A (en) | Seasonal hydrogen storage optimization configuration method of comprehensive energy system based on distributed collaborative optimization strategy | |
| CN117559526A (en) | Router-simulated energy regulation and control method based on optical storage and charging integrated charging station | |
| CN117578537A (en) | Micro-grid optimal scheduling method based on carbon transaction and demand response | |
| CN114362169A (en) | Layered coordination regulation and control method considering marginal cost of light storage type virtual power plant | |
| CN113610411A (en) | Multi-time-scale nested wind, light and water multi-energy complementary scheduling method | |
| CN109038654A (en) | A kind of distribution system optimizing operation method considering that distributed wind-powered electricity generation Thief zone is grid-connected | |
| Kumar et al. | Sensitivity analysis based multi-objective economic emission dispatch in microgrid | |
| CN115577549B (en) | Virtual power plant closed-loop direct control method containing multiple types of resources | |
| CN116742664A (en) | Short-term battery energy storage and seasonal hydrogen storage collaborative planning method and system | |
| Meng et al. | Enhancing energy efficiency in distributed systems with hybrid energy storage | |
| Liang et al. | Risk Stabilization and Market Bidding Strategy of Virtual Power Plant Alliance Based on Multi-stage Robust Optimization |
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 |