CN115622146A - A scheduling decision-making method for a cascaded water-solar-storage complementary system - Google Patents

Abstract

The invention discloses a scheduling decision-making method for a cascaded water-solar-storage-storage complementary system, which relates to the technical field of energy comprehensive utilization, including: taking photovoltaic power generation source side fluctuations, grid-connected point fluctuations, and economy as cascade water-solar-storage-storage scheduling optimization targets, and combining grid-connected points Exchange power constraints, hydropower station and pumped-storage reservoir water constraints, node voltage and feeder current constraints to build a cascade water-solar-storage scheduling model; convert the cascade water-solar-storage scheduling model into a Markov decision-making process; build a cascade water-solar energy storage system based on reinforcement learning Storage dynamic scheduling framework; take the current cascaded water-solar-solar-storage complementary system data as input, use the DDPG algorithm to solve the cascade water-solar-storage scheduling model transformed into a Markov decision process, and output a cascade water-solar-storage system that can cope with strong random photovoltaic output Real-time scheduling strategy. The invention has high calculation efficiency, greatly alleviates the fluctuation rate of the power supply side, improves the dispatching ability of the system, realizes the full consumption of photovoltaics, and can realize real-time decision-making according to the random environment of water and light.

Description

A scheduling decision-making method for a cascaded water-solar-storage complementary system

technical field

The invention relates to the technical field of energy comprehensive utilization, in particular to a scheduling decision-making method for a cascaded water-photovoltaic-storage complementary system.

Background technique

The strong randomness of photovoltaic output, the uncertainty of hydropower incoming water, and the strong coupling of cascaded hydropower output pose severe challenges to the operation of cascaded hydro-photovoltaic-storage systems. At present, research on multi-energy complementary and coordinated power generation technology at home and abroad has achieved some results. Aiming at the randomness of renewable energy output, conventional power sources with flexible adjustment capabilities, such as hydropower, thermal power, etc., or energy storage equipment, through advanced regulation The means are complementary to renewable energy, improve the utilization rate of renewable energy, realize multi-energy complementary joint power generation, and provide reference for the operation of cascaded water-photovoltaic-storage systems. However, most of the existing research focuses on the complementarity of a single large-capacity hydropower station and photovoltaics. With the cascade development of the river basin for many years, multiple cascade small hydropower stations have been gradually planned and built. Due to the limited capacity adjustment capacity of small hydropower stations, there are also run-of-the-river hydropower stations without regulation. Therefore, the existing traditional hydro-photovoltaic hybrid system power generation technology cannot be applied to this cascaded hydro-photovoltaic storage system. At the same time, most of the existing hydro-solar hybrid dispatching research focuses on hydropower generation and operation economics, taking into account the photovoltaic consumption rate, However, it does not look at the problem from the perspective of full consumption of new energy, and does not fully consider the real-time random fluctuations of photovoltaics. In addition, traditional optimal scheduling mainly focuses on long-term or day-ahead scheduling, which does not have much guiding significance for real-time scheduling decisions, and these involved scheduling decision-making methods are difficult to accurately adapt to the random dynamic changes of photovoltaics and incoming water. It is an urgent problem to be solved to carry out real-time scheduling decision-making of cascade water-photovoltaic-storage system with strong randomness.

Contents of the invention

The present invention aims to provide a scheduling decision-making method for a cascaded water-solar-storage complementary system, which can alleviate the above-mentioned problems.

In order to alleviate the above-mentioned problems, the technical scheme that the present invention takes is as follows:

The invention provides a scheduling decision-making method for a cascaded water-solar-storage complementary system, comprising the following steps:

S1. Taking photovoltaic power generation fluctuations, grid-connected point fluctuations, and economics as cascaded hydro-solar-solar-storage scheduling optimization objectives, combined with grid-connected point exchange power constraints, hydropower stations and pumped-storage reservoir water constraints, node voltage and feeder current constraints to build cascade hydro-solar storage scheduling model;

S2. Transform the cascade water-solar-storage scheduling model into a Markov decision-making process, and build a cascade water-solar-storage dynamic scheduling framework based on reinforcement learning;

S3. Under the framework of cascade hydro-solar-storage dynamic scheduling based on reinforcement learning, the data of the current cascade hydro-solar-storage complementary system is used as input, and the deep deterministic policy gradient (DDPG) algorithm is used to solve the problem and transform it into a Markov decision The cascaded water-solar-storage scheduling model of the process is output to obtain a real-time scheduling strategy for the cascade water-solar-storage system to deal with strong random photovoltaic output.

In a preferred embodiment of the present invention, in step S1, the cascaded water-solar-storage scheduling optimization method includes dividing the calculation cycle into M stages, according to the maximization of economic benefits, the minimization of fluctuations on the photovoltaic power generation side, and the minimization of fluctuations in grid-connected points Optimal construction of cascade water-solar-storage scheduling optimization objective function

Among them, F is the total target in the calculation period T, ER _t is the economic income of cascaded water-photovoltaic-storage system at time t, ΔP _source,t is the source side fluctuation measurement at time t, ΔP _t is the fluctuation measurement value of the grid-connected point within the period of Δt, β ₁ , β ₂ , and β ₃ are the weighting factors of the economic goal, the photovoltaic power generation source-side fluctuation stabilization goal, and the grid-connected point fluctuation stabilization goal, respectively.

In a preferred embodiment of the present invention, the information entropy theory is used to calculate the weight factors of the economic target, the photovoltaic power generation source-side fluctuation suppression target, and the grid-connected point fluctuation suppression target.

In a preferred embodiment of the present invention, the optimization goal of photovoltaic power generation source-side fluctuation is to minimize the fluctuation of photovoltaic output at each stage in the calculation period, and the calculation formula for calculating the fluctuation measure ΔP _source,t at time t is:

为计算周期第r阶段所设定的水光出力平均值。Among them, P _PV,t is the photovoltaic power generation output at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations,

The average value of water and light output set for the rth stage of the calculation cycle.

In a preferred embodiment of the present invention, the optimization goal of the grid-connected point fluctuation is to minimize the power fluctuation of the grid-connected point and form a schedulable outgoing curve. The calculation formula of the grid-connected point fluctuation measurement value ΔP _t within the Δt period is:

ΔP _t =(P _grid,t -P' _grid,t -(P _grid,t-1 -P' _grid,t-1 )) ² ,

Among them, P _grid,t is the interactive power between the cascade hydro-photovoltaic storage system and the external grid at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is Pumping and storage output at time t, P _grid,t is the interactive power between the cascade water-photovoltaic storage system and the external network at time t, P _grid,t-1 is the interactive power between the cascade water-photovoltaic storage system and the external network at time t-1, P _load,t is the load demand at time t, P′ _grid,t is the interactive power of grid-connected points before pumping and storage participates in regulation at time t, P′ _grid,t-1 is the interactive power of grid-connected points before pumping and storage participates in regulation at time t-1 , ΔP _t is the fluctuation measurement value of the grid-connected point during the Δt period.

In a preferred embodiment of the present invention, the economic optimization goal is to maximize the economic benefits of _cascaded water-photovoltaic-storage systems and external network transactions. The calculation formula is:

Among them, λ _t is the electricity price at time t, P _PV,t is the output of photovoltaic power generation at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is t Pumping and storing output in time, P _load,t is the load demand at time t.

In a preferred embodiment of the present invention, in step S1, the switching power constraint of the grid-connected point is

P _grid,min ≤P _grid,t ≤P _grid,max ,

Among them, P _grid,min and P _grid,max represent the minimum value and the maximum value of the transmission power of the grid-connected point respectively.

In a preferred embodiment of the present invention, the water volume constraints of hydropower stations and pumped storage reservoirs are

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 ,

Among them, V _i,t and V _PHS,t are the storage capacity of i cascade hydropower station and pumped storage at time t, V _i,max , V _PHS,max is the maximum value of i cascade hydropower station reservoir, pumped and stored water, SOC _{hydro, i,t} , SOC _PHS,t are the state of charge of the i-th cascade hydropower station and the reservoir water of the pumped-storage power station, and SOC _hydro,i,max , SOC _hydro,i,min are the state of charge of the i-th cascade hydropower station's reservoir capacity The maximum and minimum values of , SOC _PHS,max and SOC _PHS,min are the maximum and minimum values of the state of charge of the pumped-storage power station reservoir water, respectively.

In a preferred embodiment of the present invention, the 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, U _i,t is the voltage of node i at time t, I _j,t is the current of feeder j at time t, V _i,min and V _i,max are the allowable minimum value and Maximum value, I _j,min , I _j,max are the minimum and maximum allowable currents of the jth feeder line respectively.

In a preferred embodiment of the present invention, in step S3, the data of the current cascade water-solar-storage complementary system includes photovoltaic output data, load demand data, electricity price data and cascade water-power incoming water data; the current cascade water-solar-storage complementary system needs to be The data is divided into a training data set and a test data set. The training data set is used to train the cascade water-solar-solar-storage scheduling model converted into a Markov decision process, and the converged model network parameters are saved. The converged model network is used to obtain the scheduling decision results of the test data. That is, a real-time scheduling strategy for the cascaded water-solar-storage system to deal with strong random photovoltaic output.

Compared with prior art, the beneficial effect of the present invention is:

A staged fluctuation control strategy on the power source side is proposed, which avoids the inaccurate final scheduling strategy caused by the output difference between the photovoltaic rich light area and the light poor area; The scheduling model of the cascade water-solar-storage system that accommodates photovoltaics; the real-time interactive environment between deep reinforcement learning and the cascade water-solar-storage system scheduling model is designed, and the deep deterministic policy gradient (DDPG) algorithm is used to solve it. Dynamic scheduling strategy for random fluctuations of source and load; from the application point of view, this method greatly alleviates the fluctuation rate of the power source side, improves the dispatching ability of the system, meets the requirements of the power fluctuation rate index of the grid-connected point, and realizes the full consumption of photovoltaics. It has high computing efficiency and can realize real-time decision-making according to the random environment of water and light.

In order to make the above-mentioned objects, features and advantages of the present invention more comprehensible, the embodiments of the present invention will be described in detail below together with the accompanying drawings.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and thus It should be regarded as a limitation on the scope, and those skilled in the art can also obtain other related drawings based on these drawings without creative work.

Fig. 1 is a flowchart of a scheduling decision-making method for a cascaded water-solar-storage hybrid system;

Fig. 2 is a flow chart for solving dynamic dispatching of cascaded water-photovoltaic-storage system based on DDPG;

Figure 3 is a comparison chart of average fluctuations on the power supply side;

Figure 4 is a comparison chart of grid-connected point fluctuations before and after pumping and storage ginseng regulation.

detailed description

In order to make the purpose, 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 in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

Accordingly, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Please refer to Figure 1. The present invention provides a scheduling decision-making method for a cascaded water-solar-storage complementary system, specifically as follows:

S1. Build a cascade water-solar-storage-storage scheduling model, including cascade water-solar-storage scheduling optimization objective functions and constraints.

Among them, in constructing the cascade hydro-solar-storage dispatching model, aiming at the random fluctuation of photovoltaic power generation, hydropower is used as a complementary power source to stabilize photovoltaic fluctuations and improve the friendliness of photovoltaic access to the grid. Due to the limited adjustment capacity of small hydropower, coordinated control of pumping storage and cascade hydropower is used to stabilize the fluctuation of the tie line, improve the dispatchability of external delivery, and consider the economic benefits of the entire cascade hydro-photovoltaic-storage system.

1. Optimization objective function of cascaded water-solar-storage-storage scheduling

(1) Fluctuation stabilization on the power supply side

Fully considering that there are light-rich areas and light-poor areas in photovoltaics, the fluctuations on the source side cover the fluctuations caused by randomness and intermittency. The average output of the cycle is used as the benchmark as a measure of fluctuation, and the fluctuation is shown in the following formula:

Among them, PP _PV,t is the photovoltaic power generation output at time t, and P _av is the average value of hydro-photovoltaic power generation output within the calculation period.

However, the purpose of stabilizing fluctuations is to ensure that the power supply is schedulable and can be accepted by the grid. Setting it as a straight line can stabilize the output of the source side, but there are the following problems: (1) The photovoltaic output is extremely random, and the output during the day and night If the change is large, the straight-line target will definitely increase the demand for hydropower capacity; (2) In this paper, the cascade small hydropower regulation capacity is limited, and the straight-line target will inevitably lead to water abandonment at noon when the photovoltaic output power is the largest, but it cannot meet the demand at the moment when the photovoltaic output power is small. Photovoltaic fluctuations stabilize demand.

In view of the above-mentioned problems, the present invention fully considers that there are light-rich and light-poor areas in photovoltaics, and proposes a staged fluctuation control strategy, dividing the calculation cycle into M stages, each stage minimizes the fluctuation of photovoltaic output, and its fluctuation measure is proposed as follows: The stage-based fluctuation control strategy on the power supply side divides the calculation cycle into M stages, and each stage minimizes the fluctuation of photovoltaic output. The calculation formula of the fluctuation measure ΔP _source,t at time t in the calculation cycle is:

为计算周期第r阶段所设定的水光出力平均值。Among them, P _PV,t is the photovoltaic power generation output at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N represents the total number of hydropower stations,

The average value of water and light output set for the rth stage of the calculation cycle.

(2) The grid-connected point fluctuates smoothly

In order to improve the schedulability of the outgoing curve and realize the matching of source and load, the minimum power fluctuation of the grid-connected point is considered as the goal to form a dispatchable outgoing curve. The calculation formula of the fluctuation measurement value ΔP _t of the grid-connected point within the Δt period is:

ΔP _t =(P _grid,t -P' _grid,t -(P _grid,t-1 -P' _grid,t-1 )) ² ,

Among them, P _grid,t is the interactive power between the cascade hydro-photovoltaic storage system and the external grid at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is Pumping and storage output at time t, P _grid,t is the interactive power between the cascade water-photovoltaic storage system and the external network at time t, P _grid,t-1 is the interactive power between the cascade water-photovoltaic storage system and the external network at time t-1, P _load,t is the load demand at time t, P′ _grid,t is the interactive power of grid-connected points before pumping and storage participates in regulation at time t, P′ _grid,t-1 is the interactive power of grid-connected points before pumping and storage participates in regulation at time t-1 , ΔP _t is the fluctuation measurement value of the grid-connected point during the Δt period.

(3) Economy

The optimization goal of economy is to maximize the economic benefits of cascaded water-solar-storage system and external network transactions. Under the real-time electricity price mode, the calculation formula of economic income ER _t of cascade water-solar-storage system at time t is:

Among them, P _PV,t is the photovoltaic power generation output at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is the pumping and storage output at time t, P _load,t is the load demand at time t.

When constructing the optimization objective function of cascade hydro-solar-storage storage dispatching, the output stable power of the cascade hydro-solar-storage complementary system should be fully considered to ensure friendly access to the power grid. On this basis, how to obtain the maximum benefit under different runoff conditions should be considered. Therefore, according to economic benefits Maximization, minimization of photovoltaic power source side fluctuations, minimization of grid-connected point fluctuations to construct cascade water-solar-storage scheduling optimization objective function

Among them, F is the total target in the calculation period T, ER _t is the economic income of cascaded water-photovoltaic-storage system at time t, ΔP _source,t is the fluctuation measurement value at time t, ΔP _t is the measurement value of grid-connected point fluctuation within the period Δt, β ₁ , β ₂ , and β ₃ are the weighting factors of the economic target, the photovoltaic power generation source-side fluctuation stabilization target, and the grid-connected point fluctuation stabilization target, respectively.

In the present invention, the information entropy theory is used to calculate the weight factors of the economic target, the photovoltaic power generation source-side fluctuation suppression target, and the grid-connected point fluctuation suppression target.

2. Constraints

The cascade water-solar-storage scheduling model needs to meet the constraints of switching power at grid-connected points, water volume constraints of hydropower stations and pumped-storage reservoirs, node voltage and feeder current constraints.

(1) Switching power constraints at grid-connected points:

P _grid,min ≤P _grid,t ≤P _grid,max ,

Among them, P _grid,min and P _grid,max represent the minimum value and the maximum value of the transmission power of the grid-connected point respectively. The switching power P _grid,t of the grid-connected point is limited by the transmission capacity of the tie line, and it is not allowed to exceed its limit.

(2) Water volume constraints of hydropower stations and pumped storage reservoirs:

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 ,

Among them, V _i,t and V _PHS,t are the storage capacity of i cascade hydropower station and pumped storage at time t, V _i,max , V _PHS,max is the maximum value of i cascade hydropower station reservoir, pumped and stored water, SOC _{hydro, i,t} , SOC _PHS,t are the state of charge of the i-th cascade hydropower station and the reservoir water of the pumped-storage power station, and SOC _hydro,i,max , SOC _hydro,i,min are the state of charge of the i-th cascade hydropower station's reservoir capacity The maximum and minimum values of , SOC _PHS,max and SOC _PHS,min are the maximum and minimum values of the state of charge of the pumped-storage power station reservoir water, respectively.

(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, U _i,t is the voltage of node i at time t, I _j,t is the current of feeder j at time t, V _i,min and V _i,max are the allowable minimum value and Maximum value, I _j,min , I _j,max are the minimum and maximum allowable current of the jth feeder line respectively.

S2. Transform the cascade water-solar-storage scheduling model into a Markov decision-making process, and build a cascade water-solar-storage dynamic scheduling framework based on reinforcement learning. Among them, it includes the construction of actions, states, and rewards during the step-by-step water-solar-storage real-time dispatching system reinforcement learning task conversion process.

1. Action

The control center of the cascaded water-solar-storage complementary dynamic dispatching system is equivalent to an MDP agent. The agent is based on the real-time status information of the observed system environment, such as: electricity price, photovoltaic output power, load demand, cascaded hydropower and pumped storage capacity. To guide the dispatching and operation of the cascade hydro-photovoltaic-storage system based on the requirements of economy, source side and grid-connected point power fluctuation stabilization, the cascade hydropower output power p _hydro,i,t and the pumped-storage power generation/consumption power P _PHS,t are used as the 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 are the minimum value and maximum value of the i-th hydropower output respectively, P _PHS,min , P _PHS,miax are the minimum value of the pumping storage output and the maximum value. In the process of reinforcement learning, the boundary of hydropower output and pumped storage output power is limited in the action space, so there is no need to describe it with constraint conditions.

2. Status

Through the interaction mechanism with the environment, the agent gets corresponding rewards, _st is the real-time state observation information obtained through continuous interaction with the environment, and the reinforcement learning subject decides the output power of hydropower and pumped storage through the observed state information. The state of the cascade hydro-photovoltaic-storage complementary dynamic dispatching system includes the time period, electricity price, photovoltaic output in the current period, load demand, and the state of charge (State of Charge, SOC) of the cascade hydropower station and the water capacity of the pumped reservoir 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. Rewards

The cascade hydro-photovoltaic-storage dynamic scheduling model comprehensively considers the maximization of revenue and the minimization of source-side and grid-connected point fluctuations, and obtains the maximum cumulative return through trial-and-error learning by the control center. However, the optimal strategy must satisfy the constraints in the scheduling model, so the constraints need to be converted into partial rewards, which is equivalent to converting an optimization problem with constraints into an optimization problem without constraints. The reward function is expressed as follows:

r _total,t =β ₁ ER _t -β ₂ ΔP _source,t -β ₃ ΔP _t ,

Among them, δ ₁ is the penalty coefficient corresponding to the node voltage exceeding the limit value, and δ ₂ is the penalty coefficient of the branch current exceeding the allowable range. In the present invention, δ ₁ and δ ₂ are set as constants. The reservoir water volume of cascade hydropower stations and pumped-storage power stations is analogous to the state of charge SOC of the battery. δ _k,t is the corresponding penalty item when the k-th SOC exceeds the upper and lower limits, including the storage capacity SOC of hydropower stations and pumped-storage power stations, and ω is the penalty coefficient.

S3. Under the framework of cascade hydro-solar-solar-storage dynamic scheduling based on reinforcement learning, the current cascade water-solar-storage complementary system data (photovoltaic output data, load demand data, electricity price data, and cascade hydropower incoming water data) are used as input, and the current The data of the solar-storage complementary system is divided into training data sets and test data sets, and the DDPG algorithm is used to solve the cascade water-solar-storage scheduling model converted into a Markov decision process (that is, the DDPG algorithm is used to continuously try and error through multiple processes to search for an optimal scheduling strategy ), the output is the real-time scheduling strategy of the cascaded water-solar-storage system for strong random photovoltaic output, as shown in Figure 2, and the specific process is as follows:

1) Initialize the cascade water-solar-storage scheduling model parameters and DDPG network hyperparameters, weights and biases.

2) Randomly read in one day's photovoltaic output data, load demand data, electricity price data and cascade hydropower incoming water data in the training data set, update the environmental model, and obtain the initial state.

3) Based on the current strategy, the actions are cascade hydropower and pumped-storage output power. According to the constructed cascade hydro-photovoltaic-storage dynamic scheduling environment model, the immediate reward is calculated and the next moment status is output.

4) Generate a tuple (including: state, action, reward, next moment state) and store it in the experience pool, and add one to the counter of the experience pool.

5) Judging whether the experience pool is full, if it is full, select L tuples to update the strategy and value network parameters, and then continue to step 6), if not, jump to step 2);

6) Judging whether the current training is completed, that is, if ep>N, it is judged that all processes have been trained, and jump to step 7), otherwise jump to step 2);

7) Output the cumulative rewards of multiple rounds of training, observe whether it converges, if converged, save the network parameters, otherwise jump to step 1);

8) Obtain the scheduling result of the test data by using the saved model network, and output the scheduling result.

In the present invention, the model parameters converged in the training process are used to make scheduling decisions on the test data, and the results after the decision can be compared with other traditional methods, such as particle swarm and random programming methods.

The dispatching decision-making method of the cascaded water-solar-storage complementary system of the present invention has been applied in reality, and the test result analysis of 6 days is optional, and the source-side volatility and grid-connected point volatility results before and after optimization are shown in Figures 3 and 4 , although the fluctuation rate is different in different time periods, after the dynamic adjustment of cascade hydropower output, the power fluctuation on the source side has been greatly alleviated, and the average daily fluctuation rate has dropped from 32.71% to 5.97%, which is about 27% lower. After the intelligent body dynamically adjusts the operating conditions of pumped storage, the power fluctuation of the grid-connected point has been improved. The DDPG intelligent body can dynamically schedule the pumped-storage output in real time according to the current electricity price, load, and grid-connected point fluctuations, and the average fluctuation rate has been reduced from 9.03% to 6.57%. , a decrease of about 2.46%, meeting the requirement of a volatility index of less than 8%, and realizing the full consumption of photovoltaics. DDPG offline training is time-consuming, but during online testing, decisions can be made online based on the trained model, and the response can reach the second level. The solution time of the calculation example test is shown in Table 1. It can be seen that the solution time of stochastic programming and particle swarm optimization is 16.23s and 90.88s respectively, while DDPG only needs 0.62s, and the scheduling method in the present invention can achieve second-level decision-making.

Table 1 The solution time of different methods

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. A cascaded water-solar-storage complementary system dispatching decision-making method, is characterized in that, comprises the following steps: S1. Taking photovoltaic power generation fluctuations, grid-connected point fluctuations, and economics as cascaded hydro-solar-solar-storage scheduling optimization objectives, combined with grid-connected point exchange power constraints, hydropower stations and pumped-storage reservoir water constraints, node voltage and feeder current constraints to build cascade hydro-solar storage scheduling model; S2. Transform the cascade water-solar-storage scheduling model into a Markov decision-making process, and build a cascade water-solar-storage dynamic scheduling framework based on reinforcement learning; S3. Under the framework of cascade hydro-solar-storage dynamic scheduling based on reinforcement learning, the data of the current cascade hydro-solar-storage complementary system is used as input, and the deep deterministic policy gradient (DDPG) algorithm is used to solve the problem and transform it into a Markov decision The cascaded water-solar-storage scheduling model of the process is output to obtain a real-time scheduling strategy for the cascade water-solar-storage system to deal with strong random photovoltaic output. 2. The scheduling decision-making method for cascaded water-solar-storage-storage complementary system according to claim 1, characterized in that, in step S1, the cascade water-solar-storage scheduling optimization method includes maximizing economic benefits, minimizing photovoltaic power generation source side fluctuations, and Minimizing network fluctuations to construct cascaded water-solar-storage scheduling optimization objective function

Among them, F is the total target in the calculation period T, ER _t is the economic income of cascaded water-photovoltaic-storage system at time t, ΔP _source,t is the source side fluctuation measurement at time t, ΔP _t is the fluctuation measurement value of the grid-connected point within the period of Δt, β ₁ , β ₂ , and β ₃ are the weighting factors of the economic goal, the photovoltaic power generation source-side fluctuation stabilization goal, and the grid-connected point fluctuation stabilization goal, respectively. 3. The dispatching decision-making method for cascaded hydro-solar-storage complementary system according to claim 2, characterized in that the weight factors of the economical target, the fluctuation stabilization target of the photovoltaic power generation source side, and the fluctuation stabilization target of grid-connected points are calculated using the information entropy theory. 4. The cascaded hydro-solar-storage complementary system scheduling decision-making method according to claim 2, characterized in that, the optimization goal of photovoltaic power generation source side fluctuation is to minimize the fluctuation of photovoltaic output in each stage in the calculation cycle, and at time t in the calculation cycle The calculation formula of the volatility measure ΔP _source,t is:

Among them, P _PV,t is the photovoltaic power generation output at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N represents the total number of hydropower stations,

The average value of water and light output set for the rth stage of the calculation cycle. 5. The scheduling decision-making method for cascaded water-solar-storage complementary system according to claim 4, characterized in that the optimization target of the grid-connected point fluctuation is to minimize the power fluctuation of the grid-connected point to form a schedulable outgoing curve, and the grid-connected point within the Δt period The calculation formula of the fluctuation measure value ΔP _t is: ΔP _t =(P _grid,t -P' _grid,t -(P _grid,t-1 -P' _grid,t-1 )) ² ,

Among them, P _grid,t is the interactive power between the cascade hydro-photovoltaic storage system and the external grid at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is Pumping and storage output at time t, P _grid,t is the interactive power between the cascade water-photovoltaic storage system and the external network at time t, P _grid,t-1 is the interactive power between the cascade water-photovoltaic storage system and the external network at time t-1, P _load,t is the load demand at time t, P′ _grid,t is the interactive power of grid-connected points before pumping and storage participates in regulation at time t, P′ _grid,t-1 is the interactive power of grid-connected points before pumping and storage participates in regulation at time t-1 , ΔP _t is the fluctuation measurement value of the grid-connected point during the Δt period. 6. The scheduling decision-making method for cascaded water-solar-storage-storage complementary system according to claim 5, characterized in that, the optimization goal of economy is to make the cascade water-solar-storage system and external network transactions obtain the maximum economic benefits, under the real-time electricity price mode , the formula for calculating the economic benefit ER _t of the cascade water-photovoltaic-storage system at time t is:

Among them, λ _t is the electricity price at time t, P _PV,t is the output of photovoltaic power generation at time t, P _hydro,i,t is the power generation output of the i-th hydropower station at time t, N is the number of hydropower stations, P _PHS,t is t Pumping and storing output in time, P _load,t is the load demand at time t. 7. The scheduling decision-making method for cascaded water-solar-storage complementary system according to claim 6, characterized in that, in step S1, the switching power constraint of grid-connected points is P _grid,min ≤P _grid,t ≤P _grid,max , Among them, P _grid,min and P _grid,max represent the minimum value and the maximum value of the transmission power of the grid-connected point respectively. 8. The dispatching and decision-making method for the cascade water-solar-storage complementary system according to claim 7, characterized in that the water volume constraints of hydropower stations and pumped-storage reservoirs are 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 , Among them, V _i,t and V _PHS,t are the storage capacity of i cascade hydropower station and pumped storage at time t, V _i,max , V _PHS,max is the maximum value of i cascade hydropower station reservoir, pumped and stored water, SOC _{hydro, i,t} , SOC _PHS,t are the state of charge of the i-th cascade hydropower station and the reservoir water of the pumped-storage power station, and SOC _hydro,i,max , SOC _hydro,i,min are the state of charge of the i-th cascade hydropower station's reservoir capacity The maximum and minimum values of , SOC _PHS,max and SOC _PHS,min are the maximum and minimum values of the state of charge of the pumped-storage power station reservoir water, respectively. 9. The cascade water-solar-storage complementary system scheduling decision-making method according to claim 8, wherein the 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, U _i,t is the voltage of node i at time t, I _j,t is the current of feeder j at time t, V _i,min and V _i,max are the allowable minimum value and Maximum value, I _j,min , I _j,max are the minimum and maximum allowable current of the jth feeder line respectively. 10. The scheduling decision-making method for cascaded water-solar-storage-storage system according to claim 9, characterized in that in step S3, the current cascade water-solar-storage system data includes photovoltaic output data, load demand data, electricity price data and cascade hydropower Water data; the data of the current cascade water-solar-storage complementary system needs to be divided into a training data set and a test data set, and the training data set is used to train the cascade water-solar-storage scheduling model converted into a Markov decision process, and the converged model network parameters are saved. The converged model network obtains scheduling decision results for the test data.

Images