A method, device and medium for simulating sequential production considering dynamic capacity planning

Abstract

The present application discloses a method, device and medium for simulating sequential production considering capacity dynamic programming, and relates to the technical field of power system. The method includes: establishing a single-year sequential production simulation model; constructing an equivalent load support rate interval decision network based on the single-year sequential production simulation model and the CNN‑GRU network; constructing a multi-stage dynamic programming model for power capacity expansion problems based on the single-year sequential production simulation model; using the NO search algorithm to compress the state space in the multi-stage dynamic programming model to obtain a compressed state space for the dynamic programming problem; in the compressed state space, using the APO optimization algorithm and the DCS algorithm to construct a sequential production simulation method solution framework considering capacity dynamic programming, and using the sequential production simulation method solution framework to determine the rapid sequential production simulation results considering capacity dynamic programming. The present application realizes rapid simulation of sequential production.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device and medium for simulating sequential production taking into account dynamic capacity planning. Background Art

[0002] In the context of refined management of power system power capacity, there is a lock-in effect in power capacity. The power capacity put into production at a certain time will have a great impact in the future. In order to achieve economic and environmental benefits in the long term, it is necessary to plan and design the power capacity of the power system from a long-term perspective, but there are currently problems that are difficult to solve. Dynamic programming problems can efficiently solve problems with overlapping sub-problems, can handle multi-stage decision-making problems, and obtain the optimal solution through recursion and combination of sub-problems. It also has a faster running time. For the problem of power system power capacity expansion considering capacity retirement, this problem can be modeled as a dynamic programming problem with after-effects, but there are currently mathematical problems to solve dynamic programming problems with after-effects. Summary of the invention

[0003] The purpose of this application is to provide a method, device and medium for simulating sequential production taking into account dynamic capacity planning, thereby realizing rapid simulation of sequential production.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a time-series production simulation method considering dynamic capacity planning, comprising:

[0006] Establishing a single-year time series production simulation model; the single-year time series production simulation model includes: a single-year total cost objective function and a first constraint condition;

[0007] Based on the single-year time series production simulation model and the CNN-GRU network, an equivalent load support rate interval decision network is constructed;

[0008] Based on the single-year sequential production simulation model, a multi-stage dynamic programming model for the power capacity expansion problem is constructed; the multi-stage dynamic programming model includes: a dynamic programming objective function and the first constraint condition;

[0009] Using the NO search algorithm, compressing the state space in the multi-stage dynamic programming model to obtain a compressed state space of the dynamic programming problem;

[0010] In the compressed state space, the APO optimization algorithm and DCS algorithm are used to build a solution framework for the sequential production simulation method considering capacity dynamic programming.

[0011] The fast sequential production simulation result considering capacity dynamic programming is determined by using the sequential production simulation method solution framework, the equivalent load support rate interval decision network and the multi-stage dynamic programming model.

[0012] Optionally, the single-year total cost objective function includes:

[0013] F a =F inv +F main +F g +F new +F ess ;

[0014]

[0015] Among them, F a is the total cost in a single year; F inv is the investment cost; F main is the maintenance cost; F g F is the operating cost of thermal power; new Penalty cost for wind and solar power curtailment; F ess The operating cost of energy storage charging and discharging; Ω k A collection of various types of units; a k,y is the annual equivalent parameter of the unit capacity investment cost of the k-th type unit in year y; is the rated installed capacity of the kth type unit in year y; b k,y is the annual equivalent parameter of the unit capacity maintenance cost of the k-th type unit in the y-th year; Ω s is a set of running scenarios; Ω t A collection of running moments in a scene; is the unit electricity cost coefficient of thermal power units in year y; p g,s,t is the actual output of the thermal power unit at the tth moment in the sth scenario; Δt is the time interval; is the penalty cost coefficient for unit abandoned electricity of wind and photovoltaic units in year y; is the predicted output of the photovoltaic unit at the tth moment under the sth scenario; p pv,s,t is the actual output of the PV unit at the tth moment in the sth scenario; is the predicted output of the wind turbine at the tth moment under the sth scenario; p w,s,t is the actual output of the wind turbine at the tth moment in the sth scenario; is the unit power operation cost coefficient of the electrochemical energy storage system in year y; is the actual discharge power of the electrochemical energy storage system; is the actual charging power of the electrochemical energy storage system.

[0016] Optionally, the first constraint condition includes: power balance constraint, spinning reserve constraint and unit characteristic constraint;

[0017] The power balance constraint includes:

[0018]

[0019] Among them, p load,s,t is the electric load value at the tth moment in the sth scenario;

[0020] The spinning reserve constraint includes:

[0021]

[0022] Among them, R s,t is the positive spinning reserve at time t in scenario s; α max B is the maximum technical output coefficient of thermal power units; g,s,t is the grid-connected capacity of the thermal power unit at the tth moment under the sth scenario; is the rated capacity of the electrochemical energy storage system; s,t is the negative spinning reserve at time t in scenario s; α min is the minimum technical output coefficient of thermal power units; r pv is the positive reserve coefficient of the photovoltaic unit; r w is the positive reserve coefficient of the wind turbine; r load is the positive reserve factor of the electric load; o pv is the negative reserve coefficient of the photovoltaic unit; w is the negative reserve coefficient of the wind turbine; load is the negative reserve factor of the electric load;

[0023] The unit characteristic constraints include:

[0024]

[0025]

[0026]

[0027] in, is the rated capacity of the thermal power unit; p g,s,t-1 is the actual output of the thermal power unit at the t-1th moment in the sth scenario; β up is the ramp rate of the thermal power unit; β down is the ramp-down rate of the thermal power unit; is the rated capacity of the PV unit; is the rated capacity of the wind turbine; E ess,s,t is the actual energy value of the electrochemical energy storage system at the tth moment in the sth scenario; Eess,s,t-1 is the actual energy value of the electrochemical energy storage system at the t-1th moment in the sth scenario; η ess SoC is the charge and discharge efficiency coefficient of the electrochemical energy storage system; min SoC is the minimum state of charge of the electrochemical energy storage system. max is the maximum state of charge of the electrochemical energy storage system; is the rated energy of the electrochemical energy storage system; E ess,s,0 is the initial energy of the electrochemical energy storage system in the dispatching period; E ess,s,T It is the final energy of the electrochemical energy storage system during the dispatch cycle.

[0028] Optionally, based on the single-year time series production simulation model and the CNN-GRU network, an equivalent load support rate interval decision network is constructed, including:

[0029] Get the preset capacity retirement sequence;

[0030] Based on the CNN network and the GRU network, construct the CNN-GRU network;

[0031] Reconstructing the single-year time series production simulation model based on the equivalent load support rate to obtain a reconstructed single-year time series production simulation model;

[0032] Within the preset capacity retirement range, a random function is used to continuously generate the capacity retirement sequence within the planning period;

[0033] Inputting the capacity retirement sequence into the reconstructed single-year time series production simulation model to obtain the corresponding equivalent load support rate interval;

[0034] The capacity retirement sequences and the corresponding equivalent load support rate intervals are used as training data sets to train the CNN-GRU network to obtain the equivalent load support rate interval decision network.

[0035] Optionally, the dynamic programming objective function includes:

[0036]

[0037] Among them, f k (·) is the dynamic programming objective function of the kth stage; α k is the state of the kth stage; V k is the cumulative target value of the kth stage; x 0 is the decision variable in the initial stage; x 1 is the decision variable in the first stage; x k is the decision variable of the kth stage.

[0038] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described sequential production simulation methods considering dynamic capacity planning.

[0039] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned sequential production simulation methods considering dynamic capacity planning.

[0040] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned timing production simulation methods considering dynamic capacity planning.

[0041] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0042] The present application discloses a method, device and medium for simulating sequential production considering dynamic capacity programming, proposes a single-year sequential production simulation model based on a scenario method, establishes the single-year sequential production simulation model as a pure linear programming problem, has excellent solvability, and can achieve the purpose of rapid solution under a solver or a heuristic algorithm; adopts a data-model hybrid driven method to rapidly generate an equivalent load support rate interval under a capacity retirement sequence, which has the characteristics of rapidity and high precision; the present invention proposes a method for establishing a multi-stage dynamic programming model for the problem of power capacity expansion, which can use a dynamic programming method to rapidly solve the problem of power capacity expansion and improve the solution speed; a state space compression method for a dynamic programming problem based on an NO search algorithm can solve the problem of "dimensionality curse" of a dynamic programming problem, and can effectively improve the solution speed for large-scale dynamic programming problems while ensuring high precision; a solution framework for a rapid sequential production simulation method considering dynamic capacity programming, which can rapidly solve dynamic programming problems with aftereffects, adopts an APO optimization algorithm and a DCS algorithm to improve the algorithm's search and optimization capabilities, and has the characteristics of high search efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 A schematic flow chart of a sequential production simulation method considering dynamic capacity planning provided in an embodiment of the present application;

[0045] Figure 2 A schematic diagram of the framework of a fast sequential production simulation method considering dynamic capacity planning;

[0046] Figure 3 Implementing process diagram for fast sequential production simulation method considering dynamic capacity planning;

[0047] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] The purpose of this application is to provide a method, device and medium for simulating sequential production taking into account dynamic capacity planning, aiming to achieve rapid simulation of sequential production.

[0050] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0051] In an exemplary embodiment, Figure 1-Figure 3 As shown, the time-series production simulation method considering dynamic capacity planning in this embodiment includes:

[0052] Step 1: Establish a single-year time series production simulation model; the single-year time series production simulation model includes: a single-year total cost objective function and a first constraint condition.

[0053] As an optional implementation, the single-year total cost objective function includes:

[0054] F a =F inv +F main +F g +F new +F ess (1)

[0055]

[0056] Among them, F a is the total cost in a single year; F inv is the investment cost; F main is the maintenance cost; F g F is the operating cost of thermal power;new Penalty cost for wind and solar power curtailment; F ess The operating cost of energy storage charging and discharging; Ω k A collection of various types of units; a k,y is the annual equivalent parameter of the unit capacity investment cost of the k-th type unit in year y; is the rated installed capacity of the kth type unit in year y; b k,y is the annual equivalent parameter of the unit capacity maintenance cost of the k-th type unit in the y-th year; Ω s is a set of running scenarios; Ω t A collection of running moments in a scene; is the unit electricity cost coefficient of thermal power units in year y; p g,s,t is the actual output of the thermal power unit at the tth moment in the sth scenario; Δt is the time interval; is the penalty cost coefficient for unit abandoned electricity of wind and photovoltaic units in year y is the predicted output of the photovoltaic unit at the tth moment under the sth scenario; p pv,s,t is the actual output of the PV unit at the tth moment in the sth scenario; is the predicted output of the wind turbine at the tth moment under the sth scenario; p w,s,t is the actual output of the wind turbine at the tth moment in the sth scenario; is the unit power operation cost coefficient of the electrochemical energy storage system in year y; is the actual discharge power of the electrochemical energy storage system; is the actual charging power of the electrochemical energy storage system.

[0057] As an optional implementation, the first constraint condition includes: power balance constraint, spinning reserve constraint and unit characteristic constraint.

[0058] Power balance constraints, including:

[0059]

[0060] Among them, p load,s,t is the electric load value at the tth moment in the sth scenario.

[0061] Spinning reserve constraints, including:

[0062]

[0063] Among them, R s,t is the positive spinning reserve at time t in scenario s; α max B is the maximum technical output coefficient of thermal power units; g,s,t is the grid-connected capacity of the thermal power unit at the tth moment under the sth scenario; is the rated capacity of the electrochemical energy storage system;s,t is the negative spinning reserve at time t in scenario s; α min is the minimum technical output coefficient of thermal power units; r pv is the positive reserve coefficient of the photovoltaic unit; r w is the positive reserve coefficient of the wind turbine; r load is the positive reserve factor of the electric load; o pv is the negative reserve coefficient of the photovoltaic unit; w is the negative reserve coefficient of the wind turbine; load is the negative reserve factor of the electric load.

[0064] Unit characteristic constraints, including:

[0065]

[0066]

[0067]

[0068] in, is the rated capacity of the thermal power unit; p g,s,t-1 is the actual output of the thermal power unit at the t-1th moment under the sth scenario; β up is the ramp rate of the thermal power unit; β down is the ramp-down rate of the thermal power unit; is the rated capacity of the PV unit; is the rated capacity of the wind turbine; E ess,s,t is the actual energy value of the electrochemical energy storage system at the tth moment in the sth scenario; E ess,s,t-1 is the actual energy value of the electrochemical energy storage system at the t-1th moment in the sth scenario; η ess SoC is the charge and discharge efficiency coefficient of the electrochemical energy storage system; min SoC is the minimum state of charge of the electrochemical energy storage system. max is the maximum state of charge of the electrochemical energy storage system; is the rated energy of the electrochemical energy storage system; E ess,s,0 is the initial energy of the electrochemical energy storage system in the dispatching period; E ess,s,T It is the final energy of the electrochemical energy storage system during the dispatch cycle.

[0069] Specifically, the single-year time series production simulation model is established based on the scenario method: first, the historical data of wind power output, photovoltaic output and electric load are obtained, and the k-means clustering algorithm is used to generate a typical day scenario. The typical day scenario is specifically manifested as the wind power output, photovoltaic output and electric load data within a typical day. The method replaces the data of one year with the typical day scenario data. Secondly, based on the typical day scenario, a single-year time series production simulation model is established. Finally, the single-year time series production simulation model is established as a pure linear programming problem with excellent solvability, and Python is used to call the Gurobi solver for solution. The input data is the typical day scenario data and the installed capacity of the unit (the unit includes thermal power units, wind power units, photovoltaic units and electrochemical energy storage systems), and the output data is the operation status of the unit within a typical day.

[0070] Step 2: Based on the single-year time series production simulation model and the CNN-GRU network, an equivalent load support rate interval decision network is constructed.

[0071] As an optional implementation, step 2 includes:

[0072] Step 21: Obtain a preset capacity retirement sequence.

[0073] Step 22: Based on the CNN network and the GRU network, build a CNN-GRU network.

[0074] Step 23: Reconstruct the single-year time series production simulation model based on the equivalent load support rate to obtain a reconstructed single-year time series production simulation model.

[0075] Step 24: Within the preset capacity retirement range, a random function is used to continuously generate a capacity retirement sequence within the planning period.

[0076] Step 25: Input the capacity retirement sequence into the reconstructed single-year time series production simulation model to obtain the corresponding equivalent load support rate range.

[0077] Step 26: Use each capacity retirement sequence and the corresponding equivalent load support rate interval as a training data set to train the CNN-GRU network to obtain an equivalent load support rate interval decision network.

[0078] Specifically, firstly, based on the single-year time series production simulation model, the equivalent load support rate interval [δ bot ,δ top ].

[0079] Under a given capacity retirement sequence, the electricity load in the single-year time-series production simulation model is reconstructed as:

[0080]

[0081] in, is the electric load value at the tth moment in the sth scenario after a reconstruction; δ 1 is the equivalent load support rate, which is the decision variable to be optimized. The minimum value of the equivalent load support rate at this time is obtained by minimizing the total cost. bot ; With the goal of maximizing the equivalent load support rate, the high value of the equivalent load support rate at this time is obtained top .

[0082] Secondly, a method for quickly generating equivalent load support rate intervals under a capacity retirement sequence driven by a data model hybrid is established. A data-driven equivalent load support rate interval decision network is established based on a convolutional neural network-gated recurrent unit (CNN-GRU). In the first step, a random function is used to continuously generate a capacity retirement sequence within the planning period, and the capacity retirement sequence is input into the reconstructed model to obtain the equivalent load support rate interval. A large number of capacity retirement sequences and corresponding equivalent load support rate intervals are obtained to construct a training data set. In the second step, a CNN-GRU network is constructed, and a deep learning model is used to learn and imitate massive data. The CNN deep learning network shows excellent performance in feature extraction. CNN uses local connections and weight sharing to automatically extract deep feature information of the data directly from the data through alternating convolutional layers and pooling layers. Therefore, CNN is used to fully extract deep temporal information to form high-dimensional feature vector data as the mapping input of the subsequent network. Since there is a close time series relationship between capacity retirement sequences, GRU, which is good at processing high-dimensional time series feature data and has a fast computing speed, is used to construct the mapping relationship between capacity retirement sequences and equivalent load support rate interval decisions. In the CNN-GRU hybrid network, CNN is mainly responsible for extracting features, and GRU is mainly responsible for establishing the mapping relationship.

[0083] Finally, the Adam algorithm is used to train the CNN-GRU network, and its weight update formula is:

[0084]

[0085]

[0086]

[0087] Among them, ω h+1 is the network weight for the h+1th training times; ω h is the network weight at the hth training time; τ is the learning rate; is the second-order moment mean of the corrected gradient under the hth training times; θ is the smoothing coefficient; is the first-order moment mean of the corrected gradient under the hth training times; β 2 is the attenuation factor of the second-order moment mean; v h-1 is the second-order moment mean of the gradient under the h-1th training number; ▽ is the gradient operator; ω h is the network weight under the hth training times; β 2 hth power; β 1 is the attenuation factor of the first-order moment mean; m h-1 is the first-order moment mean of the gradient under the h-1th training number; β 1 The hth power.

[0088] At the same time, the CNN-GRU network is trained using a variable learning rate, that is, the learning rate decreases with the increase in the number of training times. The loss function of the model training is defined as:

[0089]

[0090] Among them, R MSE is the root mean square error; is the actual value of the h1th equivalent load support rate interval, q h1 is the estimated value of the h1th equivalent load support rate interval output by the CNN-GRU network; H is the number of equivalent load support rate intervals.

[0091] Step 3: Based on the single-year time-series production simulation model, a multi-stage dynamic programming model for the power capacity expansion problem is constructed; the multi-stage dynamic programming model includes: a dynamic programming objective function and a first constraint condition.

[0092] As an optional implementation, the dynamic programming objective function includes:

[0093]

[0094] Among them, f k (·) is the dynamic programming objective function of the kth stage; α k is the state of the kth stage; V k is the cumulative target value of the kth stage; x 0 is the decision variable in the initial stage; x 1 is the decision variable in the first stage; x k is the decision variable of the kth stage.

[0095] Specifically, when solving the problem of power capacity expansion, it is necessary to make decisions on the power capacity of each stage during the planning period. The multi-stage dynamic programming model takes the problem of power capacity expansion of the power system as the research content and uses the dynamic programming method to mathematically express the problem of power capacity expansion.

[0096] The objective function of the multi-stage dynamic programming model is the total cost, including the investment cost, maintenance cost, thermal power operation cost, wind and solar power curtailment penalty cost, and energy storage charging and discharging operation cost in each stage. The state is the installed capacity of each type of power source in each stage. The stage is the time scale for making a power capacity decision. The decision is the selection behavior from one state to the next state in the stage, specifically the installed capacity of each type of power source selected for expansion. The strategy is a sequence of decisions in each stage, that is, the strategy for expanding the capacity of each type of power source.

[0097] Characterize the power system power capacity expansion problem using dynamic programming:

[0098] 1) Stage: The planning period is divided into stages, with one year as one stage. There are Y years in total, that is, there are Y stages in the planning period, and stage k is represented as the kth stage.

[0099] 2) State: The feasible power capacity in each stage is represented as state α, which has four categories: wind power, photovoltaic power, thermal power and electrochemical energy storage. The dimension of each state vector is 4. For the representation of the state set to be selected in the stage, a fixed capacity interval is used to generate the state subset I.

[0100] 3) Decision of state transfer:

[0101] Decision variables: x k,j (α k,i ) is from α k,i Departure to α k+1,j The decision variable at the end is the capacity of the k-th stage expansion, and the following relationship exists:

[0102] α k+1,j =α k,i +x k,j (α k,i ) (14)

[0103] Among them, α k+1,j is the jth state in the k+1th stage; α k,i is the i-th state of the k-th stage.

[0104] 4) The minimum planned operating cost (investment cost, maintenance cost, thermal power operating cost, wind and solar power abandonment penalty cost, and energy storage charging and discharging operating cost) is taken as the optimal indicator function.

[0105] Indicator function form:

[0106]

[0107] V k,j =v k (α k,i ,x k,j)+V k-1,i (16)

[0108] Among them, V k,j is the cumulative target value of the jth state in the k+1th stage; v q (α q ,x q ) is the target value of the qth stage; α q is the state of the qth stage; x q is the decision variable of the qth stage; v k (α k,i ,x k,j ) is the target value of the kth stage; V k-1,i is the cumulative target value of the j-th state in the k-1th stage.

[0109] 5) The optimal objective function, i.e. the dynamic programming objective function, is:

[0110]

[0111] The input data of the multi-stage dynamic programming model are typical daily scenario data and equivalent load support rate.

[0112] Regarding the solution of the optimal objective function, within a certain state, the planned operating cost is calculated based on the single-year time series production simulation model in step 1 after correction. The power load in the corrected single-year time series production simulation model is reconstructed as:

[0113]

[0114] in, is the electric load value at the tth moment in the sth scenario after the second reconstruction; δ 2 Equivalent load support ratio in the input data for the multi-stage dynamic programming model.

[0115] With the goal of minimizing the planning operation cost, Python is used to call the Gurobi solver for solution, and then the target value of the current state of the multi-stage dynamic programming model is returned.

[0116] Step 4: Use the NO search algorithm to compress the state space in the multi-stage dynamic programming model to obtain the compressed state space of the dynamic programming problem.

[0117] Specifically, when the stage and state set of the multi-stage dynamic programming model increase, the computer's computational workload and the required storage space will increase dramatically, forming a "curse of dimensionality" problem. First, the local optimal capacity decision plan is obtained according to the minimization of the single-year expansion planning model, and finally a local optimal search path is retained. The non-monopoly (NO) search algorithm is adopted to explore and develop other possible and better search paths starting from the local optimal search path. When searching for a path, the local optimal value (the target value under the local optimal search path) is used as the benchmark. If the target value of the new search path is inferior to the local optimal value, the surrounding area of ​​the new search path is marked as unavailable space, and the marking degree and marking space are affected by the difference in target values; similarly, if the target value of the new search path is better than the local optimal value, the surrounding area of ​​the new search path is marked as available space. Finally, after the search is completed, according to the state space marking results, the advantages and disadvantages are offset, and the state space with better marking results is preferentially selected as the compressed state space of the dynamic programming problem.

[0118] The formula for the search path influence range is:

[0119]

[0120] Among them, A k,i,n is the position of the i-th state in the k-th stage under the n-th search path, A k,j,0 is the position of the jth state in the kth stage under the initial local optimal search path; σ + is the range coefficient of excellent effect; f 0 is the local optimal target value; f n is the target value of the new search path; σ - The range factor for the inferior effect.

[0121] If the label space that satisfies the above formula is Ω k-i , and its marking degree is expressed by the following formula:

[0122]

[0123] Among them, z k,i is the marking degree of the marking space of the i-th state in the k-th stage, and ρ is the marking degree coefficient.

[0124] The state space is selected according to the labeling result after iteration, and the criterion is defined as:

[0125] z k,i ≥ε(20)

[0126] Among them, ε is a constant coefficient.

[0127] The non-monopoly search algorithm is a single-solution metaphor-free algorithm that combines the advantages of exploration and development and can avoid the problem of falling into suboptimal solutions. First, the population is initialized according to the capacity of various units in the planning period under the local optimal search path; the next step is to explore according to formula (21), calculate the target value, and continuously improve the search direction by judging the size of the target value. When the number of upper-level iterations exceeds half, it is developed according to formula (22), that is, formula (22) replaces formula (21), and iterative optimization continues; finally, the NO algorithm continuously adjusts the search direction until the algorithm reaches the upper limit of the number of iterations and ends the iteration process.

[0128] X new (j) = rand * X (RP) (21)

[0129] X new (j)=X(j)-[X(SRP)*rand]*eps-[X(j)-NO](22)

[0130] Among them, X new is the newly generated capacity decision plan for each state; rand is a random value between 0 and 1; j is the dimension index of the jth state under the capacity decision plan; X is the current capacity decision plan for each state; RP and SRP are both random dimension indexes, between 1 and the maximum dimension; eps is a small value parameter; NO is an adjustment parameter.

[0131] Step 5: In the compressed state space, use the APO optimization algorithm and DCS algorithm to build a solution framework for the sequential production simulation method that considers capacity dynamic programming.

[0132] Specifically, the solution framework of the sequential production simulation method is divided into an outer stage and an inner stage. The outer stage uses an artificial protozoan optimization algorithm to generate a capacity retirement sequence, and quickly obtains the equivalent load support rate interval according to the mathematical model hybrid drive method; the inner stage uses a differentiated creative search algorithm based on the interval transmitted by the outer layer to generate an equivalent load support rate coefficient, which is output to a multi-stage dynamic programming model to obtain the cost of the expansion problem, and the retirement cost of the retirement problem is obtained based on the equivalent load support rate coefficient and the capacity retirement sequence, and then the cost of the original problem is calculated. After the inner layer iteration is completed, it is output to the outer layer until the outer layer converges. Output the fast sequential production simulation results considering capacity dynamic programming.

[0133] Artificial Protozoa Optimizer (APO) optimization algorithm: It imitates the survival behavior of protozoa in nature, including foraging, dormancy and reproduction, and has the characteristics of being simple and effective for optimization problems. Formula (23) simulates the autotrophic behavior, heterotrophic behavior, dormancy behavior and reproduction behavior of protozoa respectively. In each iteration, the probability of various behaviors of protozoa is first calculated according to formula (24) and the appropriate behavior is selected from formula (23) for iteration. Finally, the protozoa position information is judged whether it is updated according to formula (25).

[0134]

[0135]

[0136]

[0137] In formula (23), is the updated position of the e-th protozoan; W e is the original position of the e-th protozoan; f 1 W is the foraging factor; y is the yth randomly selected protozoan; N is the number of neighboring points; w 1 is the weight factor of autotrophic behavior; W γ- The original position of a protozoa with a sorting index less than e is randomly selected from the neighboring points of the kth protozoa; W γ+ The original position of a protozoa with a sorting index greater than e is randomly selected from the neighboring points of the kth protozoa; ⊙ is the Hadamard product; M f is the mapping vector of autotrophic behavior and heterotrophic behavior; W near W e Nearby location; 2 is the weight factor of heterotrophic behavior; W e-γ is the protozoa selected from the protozoa neighbors of the γth protozoa, and its ranking index is e-γ; W e+γ is the protozoa selected from the protozoa neighbors of the γth protozoa, and its ranking index is e+γ; W min is the lower bound of the protozoan position; Rand is a random vector whose elements are in the interval [0,1]; W max is the upper limit of the protozoan position; M r is the mapping vector of the reproduction behavior.

[0138] In formula (24), pf is the proportion of dormancy and reproduction in the protozoan population; pf max is the maximum value of pf; p ah is the probability of autotrophic and heterotrophic behavior, iter is the current iteration number; itermax is the upper limit of the number of iterations; p dr is the probability of dormancy and reproduction; ps is the protozoan population size.

[0139] In formula (25), is the target value of the updated position of the e-th protozoan; F(W e ) is the target value of the original position of the e-th protozoan.

[0140] Differentiated innovative search algorithm:

[0141] Differentiated creative search (DCS) is a breakthrough optimization algorithm that has completely changed the traditional decision-making system in complex environments. The main goal of DCS is to improve decision-making efficiency by adopting a dual-strategy approach that balances divergent and convergent thinking. The DCS method first performs differentiated knowledge acquisition on the randomly initialized decision variables of team members according to formula (26) to determine the degree of imperfect knowledge of team members; Formula (27) represents the iterative equations of convergent thinking and divergent thinking respectively. Combined with the current degree of imperfect knowledge of team members, the variation direction of the decision variables is further determined according to formula (27) and the updated decision variable information is obtained; finally, the decision variable information is iterated according to formulas (28) and (29) to obtain the optimal target value.

[0142]

[0143]

[0144]

[0145]

[0146] In formula (26), is the variable value of the lth member at the tth iteration; R l,t is the serial number of the lth member at the beginning of the tth iteration; NP is the number of team members.

[0147] In formula (27), o l,d is the dth element of the lth member after update; w v is the cognitive weight of the best performance; u best,d is the dth element of the best performing member in the current iteration; l,t and ω l,t are the coefficient values ​​of the lth member at the tth iteration; u r2,d is the dth element of the r2th member randomly selected in {1,2,...,NP}; u l,d is the dth element of the lth member before the update; u r1,dis the dth element of the r1th member randomly selected in {1,2,...,NP}; Lk(α v ,σ v ) is the control parameter ξ v and σ v Linique distributed random number generator.

[0148] In formula (28), D l,t+1 is the updated variable value of the lth member at the t+1th iteration; C l,t is the variable value of the lth member before updating at the tth iteration; F 2 (D l,t ) is the updated target value of the i-th member at the t-th iteration; F 2 (C l,t ) is the target value of the i-th member before updating at the t-th iteration; D l,t is the updated variable value of the lth member at the tth iteration.

[0149] In formula (29), C best,t+1 is the best performing member at the t+1th iteration; C l,t+1 is the variable value of the lth member before updating at the t+1th iteration; F 2 (C l,t+1 ) is the target value of the i-th member before updating at the t+1th iteration; F 2 (C best,t ) is the target value of the best performing member at the tth iteration; C best,t is the best performing member at the tth iteration.

[0150] Step 6: Use the sequential production simulation method solution framework, equivalent load support rate interval decision network and multi-stage dynamic programming model to determine the fast sequential production simulation results considering capacity dynamic programming.

[0151] In an exemplary embodiment, a computer device is provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the sequential production simulation method considering capacity dynamic programming in Example 1.

[0152] In an exemplary embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for simulating sequential production considering dynamic capacity planning in embodiment 1 is implemented.

[0153] In an exemplary embodiment, a computer program product is provided, including a computer program, which, when executed by a processor, implements the method for simulating sequential production considering dynamic capacity programming in embodiment 1.

[0154] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store timing production simulation data considering dynamic capacity planning. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a timing production simulation method considering dynamic capacity planning is implemented.

[0155] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0157] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0158] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0159] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0160] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for simulating sequential production considering dynamic capacity planning, characterized in that: The method comprises: Establishing a single-year time series production simulation model; the single-year time series production simulation model includes: a single-year total cost objective function and a first constraint condition; Based on the single-year time series production simulation model and the CNN-GRU network, an equivalent load support rate interval decision network is constructed; Based on the single-year sequential production simulation model, a multi-stage dynamic programming model for the power capacity expansion problem is constructed; the multi-stage dynamic programming model includes: a dynamic programming objective function and the first constraint condition; Using the NO search algorithm, compressing the state space in the multi-stage dynamic programming model to obtain a compressed state space of the dynamic programming problem; In the compressed state space, the APO optimization algorithm and DCS algorithm are used to build a solution framework for the sequential production simulation method considering capacity dynamic programming. Determine the fast sequential production simulation result considering capacity dynamic programming by using the sequential production simulation method solution framework, equivalent load support rate interval decision network and multi-stage dynamic programming model; Based on the single-year time series production simulation model and the CNN-GRU network, an equivalent load support rate interval decision network is constructed, including: Get the preset capacity retirement sequence; Based on the CNN network and the GRU network, a CNN-GRU network is constructed; The single-year time series production simulation model is reconstructed based on the equivalent load support rate to obtain the reconstructed single-year time series production simulation model; Within the preset capacity retirement range, a random function is used to continuously generate the capacity retirement sequence within the planning period; Input the capacity retirement sequence into the reconstructed single-year time series production simulation model to obtain the corresponding equivalent load support rate interval; Using each capacity retirement sequence and the corresponding equivalent load support rate interval as the training data set, the CNN-GRU network is trained to obtain the equivalent load support rate interval decision network; Among them, the Adam algorithm is used to train the CNN-GRU network, and the weight update formula is: Among them, ω h+1 is the network weight for the h+1th training times; ω h is the network weight at the hth training time; τ is the learning rate; is the second-order moment mean of the corrected gradient under the hth training times; θ is the smoothing coefficient; is the first-order moment mean of the corrected gradient at the hth training number; β2 is the attenuation factor of the second-order moment mean; v h-1 is the second-order moment mean of the gradient under the h-1th training number; is the gradient operator; ω h is the network weight under the hth training times; is the hth power of β2; β1 is the attenuation factor of the first-order moment mean; m h-1 is the first-order moment mean of the gradient under the h-1th training number; is β1 to the power of h; The loss function of the training model is: Among them, R MSE is the root mean square error; is the actual value of the h1th equivalent load support rate interval, q h1 is the estimated value of the h1th equivalent load support rate interval output by the CNN-GRU network; H is the number of equivalent load support rate intervals.

2. The method for simulating sequential production considering dynamic capacity planning according to claim 1, characterized in that: The single-year total cost objective function includes: F a =F inv +F main +F g +F new +F ess ; Among them, F a is the total cost in a single year; F inv is the investment cost; F main is the maintenance cost; F g F is the operating cost of thermal power; new Penalty cost for wind and solar power curtailment; F ess The operating cost of energy storage charging and discharging; Ω k A collection of various types of units; a k,y is the annual equivalent parameter of the unit capacity investment cost of the k-th type unit in year y; is the rated installed capacity of the kth type unit in year y; b k,y is the annual equivalent parameter of the unit capacity maintenance cost of the k-th type unit in the y-th year; Ω s is a set of running scenarios; Ω t A collection of running moments in a scene; is the unit electricity cost coefficient of thermal power units in year y; p g,s,t is the actual output of the thermal power unit at the tth moment in the sth scenario; Δt is the time interval; is the penalty cost coefficient for unit abandoned electricity of wind and photovoltaic units in year y; is the predicted output of the photovoltaic unit at the tth moment under the sth scenario; p pv,s,t is the actual output of the PV unit at the tth moment in the sth scenario; is the predicted output of the wind turbine at the tth moment under the sth scenario; p w,s,t is the actual output of the wind turbine at the tth moment in the sth scenario; is the unit power operation cost coefficient of the electrochemical energy storage system in year y; is the actual discharge power of the electrochemical energy storage system; is the actual charging power of the electrochemical energy storage system.

3. The method for simulating sequential production considering dynamic capacity planning according to claim 2, characterized in that: The first constraint condition includes: power balance constraint, spinning reserve constraint and unit characteristic constraint; The power balance constraint includes: Among them, p load,s,t is the electric load value at the tth moment in the sth scenario; The spinning reserve constraint includes: Among them, R s,t is the positive spinning reserve at time t in scenario s; α max B is the maximum technical output coefficient of thermal power units; g,s,t is the grid-connected capacity of the thermal power unit at the tth moment under the sth scenario; is the rated capacity of the electrochemical energy storage system; s,t is the negative spinning reserve at time t in scenario s; α min is the minimum technical output coefficient of thermal power units; r pv is the positive reserve coefficient of the photovoltaic unit; r w is the positive reserve coefficient of the wind turbine; r load is the positive reserve factor of the electric load; o pv is the negative reserve coefficient of the photovoltaic unit; w is the negative reserve coefficient of the wind turbine; load is the negative reserve factor of the electric load; The unit characteristic constraints include: in, is the rated capacity of the thermal power unit; p g,s,t-1 is the actual output of the thermal power unit at the t-1th moment under the sth scenario; β up is the ramp rate of the thermal power unit; β down is the ramp-down rate of the thermal power unit; is the rated capacity of the PV unit; is the rated capacity of the wind turbine; E ess,s,t is the actual energy value of the electrochemical energy storage system at the tth moment in the sth scenario; E ess,s,t-1 is the actual energy value of the electrochemical energy storage system at the t-1th moment in the sth scenario; η ess SoC is the charge and discharge efficiency coefficient of the electrochemical energy storage system; min SoC is the minimum state of charge of the electrochemical energy storage system. max is the maximum state of charge of the electrochemical energy storage system; is the rated energy of the electrochemical energy storage system; E ess,s,0 is the initial energy of the electrochemical energy storage system in the dispatching period; E ess,s,T It is the final energy of the electrochemical energy storage system during the dispatch cycle.

4. The method for simulating sequential production considering dynamic capacity planning according to claim 1, characterized in that: The dynamic programming objective function includes: Among them, f k (·) is the dynamic programming objective function of the kth stage; α k is the state of the kth stage; V k is the cumulative target value of the kth stage; x0 is the decision variable of the initial stage; x1 is the decision variable of the first stage; x k is the decision variable of the kth stage.

5. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the sequential production simulation method considering capacity dynamic programming as described in any one of claims 1 to 4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for simulating sequential production taking into account dynamic capacity programming as described in any one of claims 1 to 4 is implemented.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for simulating sequential production taking into account dynamic capacity programming as described in any one of claims 1 to 4 is implemented.

Images