A short-term unit commitment calculation method, system, medium and device

Abstract

The application discloses a short-term unit commitment calculation method, system, medium and equipment, utilizes offline trained convolutional neural network and full connection network to perform feature extraction and rule learning on unit start-stop timing information; in online calculation, an integer variable reduction strategy is adopted, and the unit start-stop scheme is adjusted based on the trained convolutional neural network, so that the difficulty of subsequent security constrained unit commitment calculation is greatly reduced; the application combines artificial intelligence and optimization methods, improves the calculation speed, and guarantees the feasibility of solving. Based on the improved XJTU-ROTS2017 test system and the actual system of a province, instance research shows that the application method significantly improves the calculation speed under the condition of guaranteeing the solving accuracy and precision.

Description

Technical Field

[0001] This invention belongs to the field of unit combination technology in the power dispatching field, specifically relating to a short-term unit combination calculation method, system, medium and equipment. Background Technology

[0002] Smart grids are an inevitable trend in power system development. With the rapid development of power systems, the Security-Constrained Unit Commitment (SCUC) problem, which forms the basis of power dispatch, will face numerous challenges. Meanwhile, with the widespread application of artificial intelligence (AI) technology in power systems, the relationship between AI and the SCUC problem will become closer and will receive increasing attention.

[0003] Against the backdrop of rapidly changing energy conditions, the Safety Constrained Unit Composition (SCUC) problem faces endless theoretical challenges. Solving it is a typical NP-hard problem with relatively low efficiency. Researching accelerated SCUC solution algorithms is of significant theoretical importance. Current accelerated methods include fully parallel stochastic optimization SCUC methods proposed for wind power, as well as methods that modify the mathematical model, reduce variables, or integrate units to accelerate large-scale SCUC solutions. In recent years, artificial intelligence has flourished and has been widely applied in power systems. Deep learning technology has achieved good results in solving practical problems in power systems such as transmission line fault diagnosis, transmission line icing identification, and power equipment defect text classification. As the foundation of distribution and dispatch, Safety Constrained Unit Composition (SCUC) will face many challenges. Traditional physical model-driven unit composition decision-making methods may not meet the actual needs of the rapidly developing power system. Therefore, scholars at home and abroad have gradually begun to explore the use of artificial intelligence to solve the SCUC problem. However, in practical applications, the complex network structure and large number of units pose significant challenges to the training of neural networks, and the accuracy and safety of artificial intelligence methods cannot be guaranteed, requiring urgent optimization and improvement. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a short-term unit combination calculation method, system, medium and equipment to address the shortcomings of the prior art, thereby solving the technical problems of long solution time and high solution complexity of large-scale SCUC models.

[0005] The present invention adopts the following technical solution:

[0006] A method for calculating short-term unit combination includes the following steps:

[0007] S1. Obtain basic technical data of the power system and establish an objective function to minimize the economic operating cost of the power system for the short-run SCUC model; then focus on the reduction of integer variables in each unit, add reasonable constraints to the relevant parameters of the units, and construct a short-run SCUC model that considers coal-fired power generation, hydropower, renewable energy and energy storage systems.

[0008] S2. Design a deep learning-based SCUC calculation method, add linear relaxation constraints to the short-term SCUC model constructed in step S1, obtain relaxation units by calculating linear problems, and then analyze the state curve of each relaxation unit through a CNN network to obtain the state probability of each relaxation unit; then modify the original start-up and shutdown scheme through a fully connected neural network, perform unit combination calculation to obtain the final scheme and output it.

[0009] Specifically, in step S1, the basic technical data of the power system includes the technical parameters of various types of power sources in the power system, renewable energy information, grid parameters, and load demand parameters.

[0010] Specifically, in step S1, the objective function of the short-term SCUC model is:

[0011]

[0012] Among them, Ω TG / Ω NEG A collection of coal-fired power units and renewable energy units; For coal-fired power units: fuel cost, start-up cost, and shutdown cost; This is an indicator variable for whether the coal-fired power unit is started / stopped; Forecasted / actual output of renewable energy; β NE is the unit cost of renewable energy waste resources; T is the number of hours included in the simulation time scale.

[0013] Furthermore, the constraints of the short-run SCUC model are as follows:

[0014] State logic constraints of thermal power units:

[0015]

[0016]

[0017] Unit ramp-up constraints:

[0018]

[0019] Minimum start-up and shutdown time constraints:

[0020]

[0021]

[0022] Where t = 1, ..., T, These are indicator variables for the status of coal-fired power units. Let be the actual power of thermal power unit i at time t. For coal-fired unit i, the maximum rate of ascent and the maximum rate of descent are given. The minimum start-up / shutdown time for coal-fired unit i;

[0023] Hydropower unit output constraints:

[0024]

[0025] Water and electricity supply constraints:

[0026]

[0027] in, Ω represents the actual power of hydropower unit i at time t. H A collection of hydroelectric generating units; Ω RH The set of watersheds contained in the system; Ω RH,k Let be the set of hydroelectric generating units contained in the k-th watershed of the system; Let be the inflow power of the k-th basin;

[0028] Operating constraints of new energy generating units:

[0029]

[0030] Among them, Ω NE A collection of new energy generating units; Let be the actual power of the new energy unit i at time t;

[0031] Energy storage charging and discharging power constraints:

[0032]

[0033] Energy storage capacity constraints:

[0034]

[0035] Energy storage cycle regulation power constraints:

[0036]

[0037] Energy storage charge / discharge state logic constraints:

[0038]

[0039] Among them, Ω TG / ΩH / Ω NE / Ω ESS / Ω TL / Ω D A collection of energy storage systems; This represents the maximum charging / discharging power of energy storage i; Let i be the charging / discharging state variable of energy storage at time t; Let i be the charging / discharging power of the energy storage device at time t; Let i be the amount of energy stored in energy storage i at time t; The charging / discharging efficiency of energy storage i; For the energy storage power adjustment cycle; The initial charge for the energy storage i-adjustment cycle;

[0040] System power balance constraints:

[0041]

[0042] Load reserve constraints:

[0043]

[0044] Cybersecurity constraints:

[0045]

[0046] Transmission section constraints:

[0047]

[0048] Where t=1,…,T,Ω TL / Ω D A collection of tie lines / loads; Let α be the power of load i at time t; RS For load reserve factor; Ω B / Ω L / Ω Sec For the set of busbars / transmission lines / transmission sections included in the system; F l,i Let be the generator output power transfer distribution factor of node k to line l.

[0049] Specifically, in step S2, the linear relaxation constraint is added as follows:

[0050]

[0051] in, Let i represent the operating state of thermal power unit i at time t. Let be the start-up state variable of thermal power unit i at time t. Let ΩTG be the shutdown state variable of thermal power unit i at time t, and let ΩTG be the set of thermal power units.

[0052] Specifically, in step S2, a scenario is generated using a time-series bootstrapping method; then, the generated scenario is used as boundary conditions, and the SCUC method is used to calculate the state of each unit as samples; finally, the samples are used to train the convolutional neural network and the fully connected neural network offline. The cross-entropy function is used as the loss function when training the neural network, as detailed below:

[0053]

[0054] Among them, w i p(x) represents the weight of each unit type in the loss function. i ) represents the actual unit status; q(x) i H(p,q) represents the predicted unit state, H(p,q) represents the loss function, and n represents the total number of units.

[0055] Specifically, in step S2, when the model has no solution, a feasible solution is found through an iterative correction mechanism, specifically as follows:

[0056] The unit continues to satisfy

[0057] Device shut-off satisfies

[0058] The unit has the following startup operation inferences:

[0059]

[0060] The following is an inference that the unit is capable of shutdown operations:

[0061]

[0062] in, Let be the start-up state variable of thermal power unit i at time t. Let be the shutdown state variable of thermal power unit i at time t. Let Ω represent the operating state of thermal power unit i at time t. TG Let t be the set of thermal power units, t be the time point, and T be the time scale, which is 24 hours.

[0063] Secondly, embodiments of the present invention provide a short-term unit combination calculation system, comprising:

[0064] The module is constructed to obtain basic technical data of the power system and to establish an objective function that minimizes the economic operating cost of the power system for the short-run SCUC model. Then, the reduction of integer variables in each unit is considered, and reasonable constraints are added to the relevant parameters of the units to construct a short-run SCUC model that considers coal-fired power generation, hydropower, renewable energy and energy storage systems.

[0065] The calculation module designs a deep learning-based SCUC calculation method, adds linear relaxation constraints to the short-term SCUC model constructed by the building module, obtains relaxed units by calculating linear problems, and then analyzes the state curve of each relaxed unit through a CNN network to obtain the state probability of each relaxed unit. Finally, it modifies the original start-up and shutdown scheme through a fully connected neural network, performs unit combination calculations to obtain the final scheme and outputs it.

[0066] Thirdly, a computer device includes 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 steps of the above-described short-term unit combination calculation method.

[0067] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described short-term unit combination calculation method.

[0068] Compared with the prior art, the present invention has at least the following beneficial effects:

[0069] A short-term unit combination calculation method is proposed. By introducing deep learning methods from artificial intelligence, this method solves the problems of long solution time and high solution complexity of large-scale SCUC models. It improves the calculation speed while ensuring calculation accuracy. Step S1 completes the collection of input data and the establishment of mathematical models, which lays the theoretical foundation for the next step and is the cornerstone of the whole method. Step S2 completes the design of the algorithm and performs calculations based on the previous step. It is the key to obtaining accurate solution results.

[0070] Furthermore, obtaining key data from power system planning departments is a prerequisite for applying this invention.

[0071] Furthermore, a short-term SCUC model considering coal-fired power plants, hydropower, renewable energy, and energy storage systems is established, which is the core model of this invention; its objective function is to minimize the economic operating cost of the power system.

[0072] Furthermore, operational constraints are set for thermal power, hydropower, new energy units, as well as energy storage and systems, and the physical characteristics of each entity in the model are described to improve the short-term SCUC model.

[0073] Furthermore, a linear relaxation module was constructed to calculate the linear problem, obtain the relaxed solution for integer variables, and use it as the input parameter for the subsequent calculation module. This is the core algorithm of this invention. This step, by linearly relaxing the original SCUC problem, obtains a relaxation unit curve that reflects the unit's state and replaces the original input parameters, greatly reducing the complexity of the input parameters.

[0074] Furthermore, a deep learning module was constructed to analyze the data provided by the linear relaxation module using a CNN network to obtain the state of each unit, which is the core algorithm of this invention. This step analyzes the state curves of the relaxation units, successively using CNN and fully connected neural networks to analyze the unit state probabilities and determine the start-up and shutdown schemes for each unit.

[0075] Furthermore, a calibration module was constructed to ensure the feasibility of the solution and provide a detailed scheduling scheme based on the results obtained from the deep learning module. This is a key step in the present invention. This step confirms and supplements the foregoing steps. If no model solution is found, this step executes an iterative calibration mechanism to gradually relax the units to find a feasible solution and provides a detailed scheduling scheme.

[0076] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0077] In summary, the method of this invention combines artificial intelligence and optimization methods. On the one hand, artificial intelligence methods are used to accelerate the solution process, and on the other hand, optimization methods are used to ensure the reasonableness of the results. Through this framework, this invention reduces most integer variables, simplifies the complexity of the problem, and can be applied to the rapid solution of complex SCUC problems.

[0078] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0079] Figure 1 This is a schematic diagram of the process of the present invention;

[0080] Figure 2 This is a schematic diagram of the deep learning module framework;

[0081] Figure 3 A flowchart of the time series bootstrapping process;

[0082] Figure 4 The diagram shows the results of the main steps of time series bootstrapping, where (a) is the original sequence, (b) is the sequence after differencing, (c) is the sequence after STL decomposition, and (d) is the regenerated sequence.

[0083] Figure 5 This is a schematic diagram of the system's grid structure;

[0084] Figure 6 The diagram shows the unit status for each typical day, where (a) is a typical spring day of SCUC, (b) is a typical spring day of DL-SCUC, (c) is a typical summer day of SCUC, (d) is a typical summer day of DL-SCUC, (e) is a typical autumn day of SCUC, (f) is a typical autumn day of DL-SCUC, (g) is a typical winter day of SCUC, and (h) is a typical winter day of DL-SCUC.

[0085] Figure 7 The diagram shows the power balance information for each typical day, where (a) is a typical day of SCUC in spring, (b) is a typical day of DL-SCUC in spring, (c) is a typical day of SCUC in summer, (d) is a typical day of DL-SCUC in summer, (e) is a typical day of SCUC in autumn, (f) is a typical day of DL-SCUC in autumn, (g) is a typical day of SCUC in winter, and (h) is a typical day of DL-SCUC in winter.

[0086] Figure 8 This is a schematic diagram of the power supply structure of a certain actual system. Detailed Implementation

[0087] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0088] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0089] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0090] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" relationship.

[0091] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0092] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0093] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0094] This invention provides a method for short-term unit combination calculation. It utilizes an offline-trained convolutional neural network and a fully connected network to extract features and learn rules from unit start-up and shutdown timing information. In online calculation, an integer variable reduction strategy is employed, and the unit start-up and shutdown scheme is adjusted based on the trained convolutional neural network, significantly reducing the difficulty of subsequent safety-constrained unit combination calculations. This invention combines artificial intelligence with optimization methods, not only improving calculation speed but also ensuring the feasibility of the solution. Case studies based on the improved XJTU-ROTS2017 test system and provincial-level actual systems demonstrate that the method of this invention significantly improves calculation speed while ensuring accuracy and precision.

[0095] Please see Figure 1 The present invention provides a short-term unit combination calculation method, comprising the following steps:

[0096] S1. Construct a short-term SCUC model that considers coal-fired power generation, hydropower, renewable energy and energy storage systems;

[0097] First, obtain basic technical data of the power system from the power system planning department, and establish an objective function to minimize the economic operating cost of the power system for the short-term SCUC model; then focus on the reduction of integer variables in each unit and add reasonable constraints to the relevant parameters of the units.

[0098] The basic technical data of the system includes: technical parameters of various types of power sources in the power system, renewable energy information, grid parameters, and load demand parameters.

[0099] The objective function of the short-run SCUC model is to minimize the economic operating cost of the power system.

[0100]

[0101] Among them, Ω TG / Ω NEG A collection of coal-fired power units and renewable energy units; For coal-fired power units: fuel cost, start-up cost, and shutdown cost; This is an indicator variable for whether the coal-fired power unit is started / stopped; Forecasted / actual output of renewable energy; β NE The unit cost of renewable energy waste resources; T is the number of hours included in the simulation time scale, which is 24 hours.

[0102] Constraints

[0103] The operating constraints of thermal power units include the state logic constraints (2) to (3), the unit ramp-up constraints (4), and the minimum start-up and shutdown time constraints (5) to (6):

[0104]

[0105]

[0106]

[0107]

[0108]

[0109] Where t = 1, ..., T, This is an indicator variable for the status of the coal-fired unit; it is 1 if the unit is online and 0 otherwise. Let be the actual power of thermal power unit i at time t; For coal-fired unit i, the maximum rate of increase and the maximum rate of decrease are given. This represents the minimum start-up / shutdown time for coal-fired unit i.

[0110] When the unit changes from the disconnected state to the connected state, the unit is limited to first climb to When changing from a powered-on state to a powered-off state, the power can only decrease to a certain level before the power is cut off.

[0111] when Unit i changes from an off state to an on state at time t, and then changes from time t to... The device should remain on, as shown in equation (5), during which all It should be restricted;

[0112] when Similarly, we can analyze that equation (6) restricts all [periods] during this time period.

[0113] when or The constraints corresponding to the variables are invalid.

[0114] The operating constraints of hydropower units include the output constraints of hydropower units (7) and the incoming water power constraints (8):

[0115]

[0116]

[0117] in, Ω represents the actual power of hydropower unit i at time t. H A collection of hydroelectric generating units; Ω RH The set of watersheds contained in the system; Ω RH,k Let be the set of hydroelectric generating units contained in the k-th watershed of the system; Let be the inflow power of the k-th basin.

[0118] Operating constraints of new energy units

[0119] New energy sources are not dispatchable, and they are treated similarly to loads in the SCUC model. The time-series output curves of new energy sources are predicted (or simulated) in the early stage and used as boundary conditions for SCUC.

[0120]

[0121] Among them, Ω NE A collection of new energy generating units; Let be the actual power of the new energy unit i at time t.

[0122] Energy storage operation constraints include energy storage charging and discharging power constraints (10), energy storage capacity constraints (11), energy storage cycle regulation capacity constraints (12), and energy storage charging and discharging state logic constraints (13):

[0123]

[0124]

[0125]

[0126]

[0127] Among them, Ω TG / Ω H / Ω NE / Ω ESS / Ω TL / Ω D A collection of energy storage solutions; This represents the maximum charging / discharging power of energy storage i; Let i be the charging / discharging state variable of energy storage at time t; Let i be the charging / discharging power of the energy storage device at time t; Let i be the amount of energy stored in energy storage i at time t; The charging / discharging efficiency of energy storage i; For the energy storage power adjustment cycle; The initial charge for the energy storage i-adjustment cycle.

[0128] System operation constraints include system power balance constraints (14), load reserve constraints (15), network security constraints (16), and transmission section constraints (17):

[0129]

[0130]

[0131]

[0132]

[0133] Where t=1,…,T,Ω TL / Ω D A collection of tie lines / loads; Let α be the power of load i at time t; RS For load reserve factor; Ω B / Ω L / Ω Sec For the set of busbars / transmission lines / transmission sections included in the system; F l,i Let be the generator output power transfer distribution factor of node k to line l.

[0134] Because there are integer variables and The SCUC problem is a mixed-integer linear programming problem, which is NP-hard and difficult to solve directly. Therefore, the following steps mainly focus on handling these integer variables to improve the computational speed of SCUC.

[0135] S2. Design a deep learning-based SCUC calculation method, including a linear relaxation module, a deep learning module, and a correction module. The linear relaxation module calculates the linear problem and obtains the relaxed solution with integer variables, which serves as the input to the deep learning module. The deep learning module analyzes the data provided by the linear relaxation module through a CNN network to obtain the state of each unit. The correction module ensures the feasibility of the solution and provides a detailed scheduling scheme based on the results obtained by the deep learning module. All the above calculation methods must be adapted to the short-term SCUC model constructed in step S1, mainly to handle the integer variables present in the model. The linear relaxation module relaxes the integer variables, which are then input into the deep learning module to obtain the state of each unit, and the correction module is then used to obtain a feasible scheduling scheme.

[0136] A power unit can only be started or shut down within a single day at most. This is based on the fact that starting or shutting down coal-fired power units takes a long time and consumes a large amount of fuel, so they cannot be started or shut down multiple times within a single day. This aligns with the principle of minimizing the number of decision variables and reducing the problem size in the optimization model.

[0137] S201. To quickly obtain approximate solutions, the original SCUC problem is first linearly relaxed, and the original complex parameter input is converted into the state curve input of the relaxed unit, so as to achieve the goal of dimensionality reduction of the input parameters.

[0138] Linear relaxation is applied to the original SCUC problem to quickly obtain an approximate solution; compared with the basic model, the following constraints also need to be added to the model:

[0139]

[0140] in, Let represent the operating state of thermal power unit i at time t, expressed as a 0-1 variable, where 0 represents the off state and 1 represents the on state. Let be the operating state variable of thermal power unit i at time t. It is 1 when the unit changes from shut down to operating, and 0 otherwise. Let Ω be the shutdown state variable of thermal power unit i at time t. It is 1 when the unit changes from start-up to shutdown, and 0 otherwise. TG It is a collection of thermal power units.

[0141] The optimal solution to a linear problem provides a lower bound for the primal problem. Due to its linear nature, the solution after linear relaxation is usually infeasible for the primal problem. (The solution is obtained from the model.) Typically, these are decimals between 0 and 1. However, considering that the model has the same constraints as the original problem, the solution provided by the linear problem also reflects the state of the unit to some extent, in addition to equation (18). It can provide boundary conditions for effectively solving mixed integer problems.

[0142] Therefore, the relaxed unit state curve This data serves as input to the deep learning module. Directly using historical data as training data would require considering historical load curves, renewable energy output curves, generator parameters, grid information, etc., significantly increasing the training difficulty. Using this method, the input data for each unit has only t dimensions, effectively reducing the dimensionality of the input parameters.

[0143] S202. Based on the relaxed units obtained in step S201, the relaxed unit state curve of each unit is analyzed using CNN to obtain the state probability of each unit. Then, the original start-up and shutdown scheme is modified to obtain the final scheme by considering the interaction between units through a fully connected neural network.

[0144] The deep learning module determines the state of each unit based on the state curves obtained from the linear relaxation module. Since the unit can only be started or stopped once a day at most, it has only four states: kept on, kept off, running, and running off. If a unit remains on or off, It should be kept in a state of 1 or 0 throughout the day; if the device has an start-up operation, it means... From 0 to 1. Similarly, if the unit has a shutdown operation, it means... The problem is changed from 1 to 0. In this way, the original problem becomes a classic classification problem in the field of artificial intelligence.

[0145] CNNs are typically used to handle classification problems. The convolutional neural network is designed with the LeNet5 architecture, which constructs two convolutional layers and two pooling layers. After convolution and pooling, a fully connected operation is performed to output the probability corresponding to each unit state.

[0146] CNNs do not consider the mutual influence between units; they determine the unit's state solely based on the relaxed start-stop state curves. For example, if a unit's relaxed state curve is {0.01, 0.01, ..., 0.01}, the unit is likely in a normally off state. If the relaxed state curve is {0.9, 0.9, ..., 0.9}, the unit is likely in a normally open state. However, if a unit's relaxed state curve is {0.5, 0.5, ..., 0.5}, the state may not be directly determined by the convolutional neural network. This requires further consideration of the mutual influence between units through fully connected layers to determine the unit's state.

[0147] After the convolutional neural network, the model still requires a fully connected neural network to consider the state influence between different units, thus obtaining the corrected result. Since the convolutional neural network only judges the unit's state based on the relaxed start-stop state curves, its judgment of the unit's state is not comprehensive enough, and the state judgment of many units is quite ambiguous. The fully connected neural network is mainly used to handle these units in intermediate states. It determines the unit's state by considering the states of other units and the net load curve.

[0148] Please see Figure 2 First, a convolutional neural network is used to analyze the relaxation state curves of each unit to obtain the state probability of each unit. Then, a fully connected neural network is used to consider the interaction between units. The state probabilities of each unit and the net load curve obtained from the convolutional neural network are used as inputs to modify the original start-up and shutdown scheme, resulting in the final start-up and shutdown scheme for each unit.

[0149] The model adopts an offline learning training mode. First, a large number of scenarios are generated through time series bootstrapping; then, these scenarios are used as boundary conditions, and the traditional SCUC method is used to calculate the state of each unit as samples; finally, these samples are used to train the convolutional neural network and the fully connected neural network offline.

[0150] Considering that the amount of historical data may not be sufficient to support the training of the neural network, in order to generate more scenarios, the model adopts a time series bootstrap approach to process the historical output curves of photovoltaic, wind power and load, generating multiple similar curves while preserving the original time series data features.

[0151] Please see Figure 3The original time series is decomposed into a trend term, a periodic term, and a remainder term. The remainder term is then bootstrapped to generate several bootstrap remainder terms, which are then summed with the trend term and the periodic term to obtain several new time series samples. This method can expand the sample diversity and also solve the problem of needing a large number of data samples for neural network training.

[0152] When training the neural network, the model uses the cross-entropy function as the loss function. Its calculation formula is shown below:

[0153]

[0154] Among them, w i p(x) represents the weight of each unit type in the loss function. i ) represents the actual unit status; q(x) i () represents the predicted unit status.

[0155] Since the unit is usually kept in the normally open or normally closed state and rarely performs power-on or power-off actions, the weight of the open and close states needs to be increased during training.

[0156] Please see Figure 4 , Figure 4 The four figures respectively represent the original sequence, the differencing sequence, the STL decomposition sequence, and the regenerated sequence during the time series bootstrapping process. By processing the historical output curves of wind, solar, and load data, multiple similar curves were generated while preserving the characteristics of the original time series data. These curves greatly expand the sample diversity, supporting the training of the neural network and further accelerating the optimization solution process.

[0157] S203. In order to ensure that a feasible solution is obtained in step S202, when the model has no solution, a feasible solution is found through an iterative correction mechanism, and a detailed scheduling scheme is provided.

[0158] To ensure the feasibility of the solution and provide a detailed scheduling plan, an iterative correction mechanism is designed in the correction module. If no solution is found in the model, the unit needs to be relaxed, kept in an on or off state, and attempts should be made to allow shutdown or startup operations. Through iteration, the constraints on the unit state are gradually released until a feasible solution is found. Theoretically, if all constraints regarding the unit state are released, the solution calculated by the post-correction module must be consistent with the result obtained by directly solving the SCUC model.

[0159] The SCUC model mentioned here needs to include constraints on variables related to reducing coal-fired power units.

[0160] If the unit continues, equation (20) must be satisfied:

[0161]

[0162] If the device is shut down, equation (21) must be satisfied:

[0163]

[0164] If the unit has a startup operation capability, the following inference can be made:

[0165]

[0166] Similarly, if the unit has a shutdown operation, the following inference can be made:

[0167]

[0168] in, Let be the operating state variable of thermal power unit i at time t. It is 1 when the unit changes from shut down to operating, and 0 otherwise. Let be the shutdown state variable of thermal power unit i at time t. It is 1 if the unit changes from start-up to shutdown, and 0 otherwise. Let Ω represent the operating state of thermal power unit i at time t, a 0-1 variable where 0 represents the off state and 1 represents the on state. TG Let t be the set of thermal power units, t be the time point, and T be the time scale, which is 24 hours.

[0169] In this invention, detailed scheduling information is not given directly through artificial intelligence, but through traditional optimization methods. This ensures that the solution must meet the model's constraints, and also reduces the difficulty of training the model.

[0170] In another embodiment of the present invention, a short-term unit combination calculation system is provided. This system can be used to implement the above-mentioned short-term unit combination calculation method. Specifically, the short-term unit combination calculation system includes a construction module and a calculation module.

[0171] The module includes a construction module that acquires basic technical data of the power system and establishes an objective function to minimize the economic operating cost of the power system for the short-run SCUC model. Then, it focuses on reducing the integer variables in each unit, adds reasonable constraints to the relevant parameters of the units, and constructs a short-run SCUC model that considers coal-fired power generation, hydropower, renewable energy and energy storage systems.

[0172] The calculation module designs a deep learning-based SCUC calculation method, adds linear relaxation constraints to the short-term SCUC model constructed by the building module, obtains relaxed units by calculating linear problems, and then analyzes the state curve of each relaxed unit through a CNN network to obtain the state probability of each relaxed unit. Finally, it modifies the original start-up and shutdown scheme through a fully connected neural network, performs unit combination calculations to obtain the final scheme and outputs it.

[0173] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a short-term unit combination calculation method, including:

[0174] This paper acquires basic technical data of the power system and establishes an objective function to minimize the economic operating cost of the power system for a short-run SCUC model. Then, focusing on the reduction of integer variables in each unit, reasonable constraints are added to the relevant parameters of the units to construct a short-run SCUC model considering coal-fired power plants, hydropower, renewable energy, and energy storage systems. A deep learning-based SCUC calculation method is designed to add linear relaxation constraints to the short-run SCUC model. Relaxed units are obtained by calculating linear problems, and then the state curve of each relaxed unit is analyzed using a CNN network to obtain the state probability of each relaxed unit. Finally, a fully connected neural network is used to modify the original start-up and shutdown scheme, perform unit combination calculations to obtain the final scheme, and output it.

[0175] In another embodiment of the present invention, a storage medium is also provided, specifically a computer-readable storage medium (memory). This computer-readable storage medium is a memory device in a terminal device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0176] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the short-term unit combination calculation method in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps:

[0177] This paper acquires basic technical data of the power system and establishes an objective function to minimize the economic operating cost of the power system for a short-run SCUC model. Then, focusing on the reduction of integer variables in each unit, reasonable constraints are added to the relevant parameters of the units to construct a short-run SCUC model considering coal-fired power plants, hydropower, renewable energy, and energy storage systems. A deep learning-based SCUC calculation method is designed to add linear relaxation constraints to the short-run SCUC model. Relaxed units are obtained by calculating linear problems, and then the state curve of each relaxed unit is analyzed using a CNN network to obtain the state probability of each relaxed unit. Finally, a fully connected neural network is used to modify the original start-up and shutdown scheme, perform unit combination calculations to obtain the final scheme, and output it.

[0178] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0179] Case Analysis

[0180] The effectiveness of the proposed algorithm was verified using the XJTU-ROTS2017 testing system and a real-world system. 36,500 typical scene results were generated based on historical data from model training using time-series bootstrapping techniques. The convolutional kernel size was set to (1,5), the cross-entropy function was used as the loss function, and the fully connected network consisted of three equal-sized hidden layers.

[0181] This invention uses an Intel(R) Core(TM) i7-9700 @ 3.60GHz computer processor for online computation and a GeForce RTX 2060 for offline training of the neural network.

[0182] The testing system includes renewable energy, coal-fired power, hydropower, and energy storage systems, comprising 44 nodes and 32 thermal power units. After the upgrade, the installed capacity of renewable energy will be 7000MW, and the installed capacity of thermal power will be 10400MW, with renewable energy accounting for over 40% of the total installed capacity. The system grid structure diagram is shown below. Figure 5 As shown in the figure. On the test set, the accuracy of the convolutional neural network in identifying unit startup and shutdown was 99.38%. This invention calculated 365 scenarios for a certain level year, and the results are shown in Tables 1 and 2.

[0183] Table 1 Calculation Results

[0184]

[0185]

[0186] Table 2 Key Indicators of the Operation Simulation

[0187]

[0188] As shown in Tables 1 and 2, the average error of DL-SCUC is only 0.04%, which is very close to the solution of SCUC. In addition, DL-SCUC reduces the total computation time for 365 scenarios from 249.72s to 89.36s. Compared with traditional optimization algorithms, this method can guarantee higher computational accuracy and improve computational speed.

[0189] Please see Figure 6 and Figure 7 The image shows the computational results of the proposed DL-SCUC and the SCUC under typical seasonal scenarios. The left side shows the computational results of the traditional SCUC model, and the right side shows the computational results of the proposed DL-SCUC. The selected scenarios are March 15th, June 15th, September 15th, and December 15th. These scenarios represent the system's four seasons. Figure 5In the diagram, darker colors indicate the device is online, and lighter colors indicate the device is offline. It can be seen that the results of SCUC and DL-SCUC are essentially the same. The error of DL-SCUC is well within acceptable limits.

[0190] Please see Figure 8 This invention also conducted a computational test on a provincial-level actual system in a certain level year. In this provincial-level actual system, the installed capacity of wind power reached 28,338 MW, and the installed capacity of photovoltaic power reached 34,605 ​​MW, as detailed below. The installed capacity of renewable energy units accounted for more than 50%. The number of thermal power units reached 173, and the number of busbars reached 860, which poses a huge challenge to SCUC calculation.

[0191] After a year of comparative calculations across 365 scenarios, the simulation results are shown in Tables 3 and 4:

[0192] Table 3 Calculation results of actual system examples

[0193]

[0194] Table 4 Key Indicators of Actual System Simulation

[0195]

[0196] As can be seen from the table above, the method of this invention reduces the total simulation time from 16431.60s to 4453.89s while maintaining a certain level of accuracy, and still performs well in complex systems. In summary, the method of this invention combines artificial intelligence and optimization methods, significantly improving the computational speed while ensuring the accuracy and precision of the solution.

[0197] In summary, this invention provides a method, system, medium, and equipment for short-term unit combination calculation. When solving large-scale SCUC problems, it introduces deep learning methods from artificial intelligence to design a calculation method that maximizes computational speed while maintaining high computational accuracy. Through convolutional neural networks (CNNs) and fully connected neural networks, data features are extracted to derive unit startup and shutdown schemes, significantly reducing integer variables and the difficulty of subsequent calculations. The solution of the linearized SCUC is used as input to the CNN to ensure model accuracy and reduce the difficulty of neural network training. This invention combines artificial intelligence and mathematical optimization and designs a correction mechanism to ensure the feasibility of the solution.

[0198] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0199] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0200] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0201] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0202] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0203] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0204] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0205] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0206] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for calculating short-term unit combination, characterized in that, Includes the following steps: S1. Obtain basic technical data of the power system and establish an objective function to minimize the economic operating cost of the power system for the short-run SCUC model; then focus on reducing the integer variables in each unit, add reasonable constraints to the relevant parameters of the units, and construct a short-run SCUC model that considers coal-fired generators, renewable energy and energy storage systems. S2. Design a deep learning-based SCUC calculation method, add linear relaxation constraints to the short-term SCUC model constructed in step S1, obtain relaxed units by calculating linear problems, and then analyze the state curves of the relaxed units. As input to the deep learning module, the state curve of each relaxation unit is analyzed by a convolutional neural network to obtain the state probability of each relaxation unit. The state probabilities of each unit and the net load curve obtained by the convolutional neural network are used as input to a fully connected neural network. The fully connected neural network considers the interaction between units, modifies the original start-up and shutdown scheme, and obtains and outputs the final start-up and shutdown scheme of each unit. Add the following linear relaxation constraint: in, Let i represent the operating state of thermal power unit i at time t. Let be the start-up state variable of thermal power unit i at time t. Let be the shutdown state variable of thermal power unit i at time t. For a collection of thermal power units, T is the time scale, T = 24 hours; Scenarios are generated using a time-series bootstrapping method. These scenarios are then used as boundary conditions, and the SCUC method is employed to calculate the status of each unit, which serves as samples. Finally, these samples are used to train convolutional neural networks and fully connected neural networks offline. The cross-entropy function is used as the loss function during neural network training, as detailed below: in, This represents the weight of each unit type in the loss function; This reflects the actual status of the generating unit; For the predicted unit status, For loss function, This represents the total number of generating units.

2. The short-term unit combination calculation method according to claim 1, characterized in that, In step S1, the basic technical data of the power system includes the technical parameters of various types of power sources in the power system, renewable energy information, grid parameters, and load demand parameters.

3. The short-term unit combination calculation method according to claim 1, characterized in that, In step S2, when the model has no solution, a feasible solution is found through an iterative correction mechanism, specifically as follows: Unit continues, satisfying ; The device is closed, satisfying the requirements. ; The unit has the following startup operation inferences: The following is an inference that the unit is capable of shutdown operations: in, Let be the start-up state variable of thermal power unit i at time t. Let be the shutdown state variable of thermal power unit i at time t. Let i represent the operating state of thermal power unit i at time t. A collection of thermal power units. For time moments, The time scale is 24 hours.

4. A short-term unit combination calculation system, characterized in that, include: The module is constructed to obtain basic technical data of the power system and to establish an objective function that minimizes the economic operating cost of the power system for the short-run SCUC model. Then, the reduction of integer variables in each unit is considered, and reasonable constraints are added to the relevant parameters of the units to construct a short-run SCUC model that considers coal-fired power generation, renewable energy and energy storage systems. The computation module designs a deep learning-based SCUC computation method, adding linear relaxation constraints to the short-term SCUC model constructed by the building module. Relaxation units are obtained by calculating linear problems. Relaxed unit state curve As input to the deep learning module, the state curve of each relaxation unit is analyzed by a convolutional neural network to obtain the state probability of each relaxation unit. The state probabilities of each unit and the net load curve obtained by the convolutional neural network are used as input to a fully connected neural network. The fully connected neural network considers the interaction between units, modifies the original start-up and shutdown scheme, and obtains and outputs the final start-up and shutdown scheme of each unit. Add the following linear relaxation constraint: in, Let i represent the operating state of thermal power unit i at time t. Let be the start-up state variable of thermal power unit i at time t. Let be the shutdown state variable of thermal power unit i at time t. For a collection of thermal power units, T is the time scale, T = 24 hours; Scenarios are generated using a time-series bootstrapping method. These scenarios are then used as boundary conditions, and the SCUC method is employed to calculate the status of each unit, which serves as samples. Finally, these samples are used to train convolutional neural networks and fully connected neural networks offline. The cross-entropy function is used as the loss function during neural network training, as detailed below: in, This represents the weight of each unit type in the loss function; This reflects the actual status of the generating unit; For the predicted unit status, For loss function, This represents the total number of generating units.

5. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1, 2, or 3.

6. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including steps for performing the method of any of claims 1, 2, or 3.

Images