Power system production simulation method and system based on support vector machine

Abstract

The present invention relates to the technical field of power system planning, and specifically to a power system production simulation method and system based on a support vector machine, the method comprising: obtaining an operating day load forecast curve and a photovoltaic output forecast curve, and constructing a residual load feature vector and a difference matrix; based on the residual load feature difference matrix, constructing a residual load aggregatability matrix of each operating day using a support vector machine; clustering and iterating the operating days based on the residual load aggregatability matrix to extract typical days; with the goal of minimizing the total operating cost, solving the unit combination model according to the typical day residual load curve to obtain a typical day start-up and shutdown result; based on the typical day start-up and shutdown result, solving the unit combination model again according to the operating day residual load curve to obtain the output results of each period of the operating day. Its purpose is to improve the calculation efficiency of power system production simulation and reduce calculation errors.

Description

Technical Field

[0001] The present invention relates to the technical field of power system planning, and in particular to a power system production simulation method and system based on a support vector machine. Background Art

[0002] In the field of power system planning and operation analysis, production simulation is an indispensable basic tool. Its core is to carry out power balance calculations for the power system in a long-term and fine time scale under the premise of strictly following the constraints related to the operation of the unit. However, traditional time-series production simulations are often implemented with months or even years as the time span. The model contains a large number of integer variables and complex constraints, which makes the calculation time required for direct solution extremely long.

[0003] With the acceleration of the global energy transformation process, the position of renewable energy in the global power system is becoming increasingly important. Among them, photovoltaic power generation, as a key component, is showing a rapid development trend. However, the volatility and uncertainty of photovoltaic output make the constraints of production simulation more complicated and the calculation difficulty significantly increased. Existing research mainly focuses on two levels: modeling and solution. At the modeling level, researchers use unit clustering, time period aggregation, and load clustering to reduce decision variables and speed up the solution process; at the solution level, they use heuristic methods such as hot start, induced function, and neighborhood search to reduce the amount of calculation or accelerate convergence.

[0004] At present, the existing technology mainly uses two methods for load clustering in production simulation. The first method processes the load in segments, aggregates the time periods with relatively slow load change trends into an equivalent period, and assumes that the load demand can be met without changing the start and stop status of the unit. The second method aggregates the operating days with similar load characteristics into typical days for solution through clustering algorithms or heuristic algorithms, thereby reducing the number of integer variables in the model. However, the first method ignores the impact of load volatility within the segment on the solution results, which is prone to large errors. At the same time, the aggregation of loads in adjacent time periods is difficult to significantly reduce the number of integer variables, and has limited effect on improving the solution speed of the unit. The second method does not include photovoltaic output characteristics in the consideration of load characteristics, and cannot fully take into account the impact of photovoltaic output volatility and uncertainty on load clustering. It also directly equates the calculation results of typical days with the calculation results of aggregated operating days, which will also lead to large errors.

[0005] In summary, under the premise of ensuring the accuracy of production simulation results, how to effectively improve the calculation efficiency, especially how to improve the load clustering method, fully consider the impact of the volatility and uncertainty of photovoltaic output on load clustering, reduce calculation errors, and significantly reduce the number of integer variables to increase the solution speed, has become a key technical problem that needs to be solved urgently in the current field of power system production simulation. Summary of the invention

[0006] In order to improve the computational efficiency of power system production simulation and reduce computational errors, the present invention provides a power system production simulation method and system based on support vector machine, and the technical solutions adopted are as follows:

[0007] The technical solution of the first aspect of the present invention provides a power system production simulation method based on a support vector machine, the method comprising:

[0008] Obtain the load forecast curve and photovoltaic output forecast curve for the operating day, and construct the residual load characteristic vector and difference matrix;

[0009] Based on the residual load characteristic difference matrix, the residual load aggregability matrix of each operating day is constructed using support vector machine.

[0010] Based on the residual load aggregatability matrix, the operation days are clustered iteratively to extract typical days;

[0011] With the goal of minimizing the total operating cost, the unit commitment model is solved according to the typical day residual load curve to obtain the typical day start-up and shutdown results;

[0012] Based on the start-up and shutdown results of typical days, the unit combination model is solved again according to the residual load curve of the operating day to obtain the output results of each period of the operating day.

[0013] Furthermore, the load forecast curve and photovoltaic output curve of the operating day are obtained, and the residual load characteristic vector and difference matrix are constructed, including:

[0014] According to the load forecast curve and photovoltaic output curve on the operating day, the load characteristic points and photovoltaic output characteristic points are extracted;

[0015] Construct a residual load characteristic vector based on the load characteristic points and the photovoltaic output characteristic points;

[0016] The residual load difference matrix is ​​constructed based on the residual load eigenvector.

[0017] Furthermore, based on the residual load characteristic difference matrix, the residual load aggregability matrix of each operating day is constructed using support vector machine, including:

[0018] The difference of residual load feature vectors between operation days is used as input and the aggregability of residual load curves is used as output to train the support vector machine;

[0019] The trained support vector machine is used to generate the residual load aggregatability matrix based on the residual load difference matrix.

[0020] Furthermore, the operation days are clustered and iterated based on the residual load aggregatability matrix, and typical days are extracted, including:

[0021] Based on the aggregability of the operating day, the target operating day is selected from the residual load aggregability matrix for aggregation to obtain the typical day residual load curve;

[0022] Remove target operating days from the residual load aggregatability matrix and add typical days.

[0023] Furthermore, based on the aggregability of the operating day, the target operating day is selected from the residual load aggregability matrix for aggregation to obtain the typical day residual load curve, including:

[0024] The two operating days with the largest aggregability are selected for aggregation. After aggregation, the residual load curve of the typical day is obtained, which can be expressed as:

[0025]

[0026] In the formula, Indicates a typical day The residual load curve of Indicates the operating day The residual load curve of Indicates the operating day The residual load curve.

[0027] Furthermore, with the goal of minimizing the total operating cost, the unit commitment model is solved according to the typical day residual load curve to obtain the typical day start-up and shutdown results, including:

[0028] According to the typical day residual compliance curve, the objective function is constructed with the goal of minimizing the unit power generation cost and startup cost, and the unit combination model is solved based on the preset constraints.

[0029] Furthermore, the expression of the objective function is:

[0030]

[0031] In the formula, For a typical day Medium Unit In the period The power generation cost is In the period contribution Function of is the total number of units; For a typical day Medium Unit In the period Start-up costs; The number of periods in a day.

[0032] Furthermore, the preset constraints include:

[0033] The load balance constraint is configured as follows: the total output of thermal power units during the target period of a typical day is equal to the total system load during that period minus the wind power output;

[0034] The system positive reserve constraint is configured such that the sum of the maximum outputs of the thermal power units is greater than or equal to the sum of the system net load and the system positive reserve capacity;

[0035] The system negative reserve constraint is configured so that the minimum total output of thermal power units is no greater than the system net load plus the system negative reserve capacity;

[0036] The upper and lower limits of the unit output are configured so that the output of the thermal power unit is between the minimum technical output and the maximum technical output;

[0037] The unit ramp rate constraint is configured to ensure that the unit output increases or decreases within a preset rate range;

[0038] The minimum continuous start and stop time constraint of the unit is configured to maintain the preset continuous operation or shutdown time after the unit is started or shut down.

[0039] Furthermore, based on the typical day start-up and shutdown results, the unit combination model is solved again according to the residual load curve of the operating day to obtain the output results of each period of the operating day, including:

[0040] Based on the correspondence between typical days and operating days, the start and stop status of thermal power units in each period of typical days is applied to the corresponding operating days;

[0041] According to the residual load curve of the operating day, the unit combination model is solved again in combination with the objective function and the preset constraints to obtain the output results of the thermal power units at different times of each operating day.

[0042] The technical solution of the second aspect of the present invention provides a power system production simulation system based on a support vector machine, using the power system production simulation method based on a support vector machine described in the technical solution of the first aspect of the present invention, the system comprises:

[0043] A data acquisition module configured to acquire a load forecast curve and a photovoltaic output forecast curve on an operation day;

[0044] a feature construction module configured to construct a residual load feature vector and a difference matrix according to a load forecast curve and a photovoltaic output forecast curve;

[0045] an aggregatability analysis module configured to construct a residual load aggregatability matrix for each operating day using a support vector machine based on a residual load characteristic difference matrix;

[0046] A clustering module configured to iteratively cluster the operating days based on the residual load aggregatability matrix and extract typical days;

[0047] The first solution module is configured to solve the unit commitment model according to the typical day residual load curve with the goal of minimizing the total operating cost, and obtain the typical day start-up and shutdown results;

[0048] The second solving module is configured to solve the unit combination model again based on the typical day start-up and shutdown results and the residual load curve of the operating day to obtain the output results of each period of the operating day.

[0049] The present invention has the following beneficial effects:

[0050] The power system production simulation method based on support vector machine provided by the present invention constructs residual load characteristic vectors and difference matrices through operating day load prediction curves and photovoltaic output prediction curves, effectively describes the load and photovoltaic characteristics of each operating day, realizes data dimension reduction, and provides convenience for load clustering; evaluates the aggregatability of operating days through the SVM model and combines the hierarchical clustering method to avoid the influence of artificial selection of cluster number and initial point on clustering results; extracts the start and stop results of computer groups on typical days, greatly reduces model variables, shortens calculation time, and then uses the results to solve the output of each operating day unit, reduces the complexity of the unit, and finally obtains more accurate model simulation results, which has an important optimization and improvement effect on power system production simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 A method flow chart of a power system production simulation method based on a support vector machine provided by an embodiment of the present invention;

[0053] Figure 2 A schematic diagram of the structure of a power system production simulation system based on a support vector machine provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a power system production simulation method and system based on a support vector machine proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0055] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0056] The specific scheme of a power system production simulation method and system based on support vector machine provided by the present invention is described in detail below with reference to the accompanying drawings.

[0057] See also Figure 1 , which shows a method flow chart of a power system production simulation method based on a support vector machine provided by an embodiment of the present invention, the method comprising:

[0058] Step S100: Obtaining the load forecast curve and the photovoltaic output forecast curve for the operating day, and constructing the residual load characteristic vector and difference matrix;

[0059] Step S100 specifically includes:

[0060] Step S110: Extract the load characteristic points and photovoltaic output characteristic points based on the load forecast curve and photovoltaic output curve on the operating day; specifically, in the power system, the load and photovoltaic output show different changing patterns within a day. By selecting representative characteristic points to describe the key features of these curves, the amount of data can be effectively reduced and important information can be retained. Each operating day has a load forecast curve and a photovoltaic output prediction curve ,in , represents 365 operating days in a year; this embodiment selects the load forecast values ​​corresponding to the starting point, morning trough point, noon peak point, noon trough point, evening peak point, evening trough point and evening peak point in the load forecast curve of each operating day as load characteristic points, which are respectively recorded as ; The starting time, end time and peak output in the photovoltaic output prediction curve of each operating day are selected as photovoltaic output characteristic points, which are recorded as ;

[0061] Construct the residual load curve for each operating day, which can be expressed as:

[0062]

[0063] In the formula, Indicates The residual load curve of an operating day reflects the change of the power load actually required to be provided by other power sources (such as traditional generators) on that operating day over time after taking into account the photovoltaic output;

[0064] Step S120: construct a residual load feature vector according to the load feature points and the photovoltaic output feature points; specifically, the extracted load feature points and photovoltaic output feature points are combined into a vector to comprehensively describe the residual load characteristics of each operating day, so as to facilitate comparison and classification of the residual load conditions on different operating days, which can be expressed as:

[0065]

[0066] In the formula, Indicates The residual load characteristic vector of the operating day contains the load characteristic point and photovoltaic processing characteristic point information of the operating day;

[0067] Step S130: construct a residual load difference matrix according to the residual load characteristic vector; specifically, by calculating the difference between the residual load characteristic vectors of different operating days, the residual load difference matrix is ​​constructed to reflect the difference degree of the residual load characteristics of any two operating days, which can be expressed as:

[0068]

[0069] In the formula, is a ten-dimensional vector, representing the and The difference of the residual load characteristic vectors of the two operating days; each dimension corresponds to the difference between the two operating days at the corresponding characteristic points. The similarity or difference of the residual load characteristics between two operating days can be measured;

[0070] In summary, step S100 accurately extracts the load and photovoltaic output feature points from the operating day load forecast curve and the photovoltaic output forecast curve, constructs the residual load curve, and then combines the residual load feature vector and calculates the difference matrix, effectively capturing the key characteristics of the load and photovoltaic output in the power system and their relationship. It not only greatly reduces the amount of data, but also retains important information, which helps to improve the accuracy and efficiency of load clustering in power system production simulation.

[0071] Step S200: Based on the residual load feature difference matrix, the support vector machine is used to construct the residual load aggregatability matrix of each operating day; specifically, the powerful classification and regression capabilities of SVM are utilized to enable it to be trained to judge the aggregatability of their residual load curves according to the differences in the residual load feature vectors between operating days, thereby providing a basis for subsequent clustering of operating days.

[0072] Step S200 specifically includes:

[0073] Step S210: train a support vector machine with the difference in residual load characteristic vectors between operating days as input and the aggregability of the residual load curve as output; specifically, the support vector machine is a supervised machine learning algorithm, which is used for regression tasks in this embodiment, with the purpose of allowing the SVM to learn the mapping relationship between the difference in residual load characteristic vectors between operating days and the aggregability of the residual load curve. Through training with a large amount of data with known aggregability labels, the SVM can autonomously judge the aggregability of the residual load curves of any two operating days. In this embodiment, the difference in residual load characteristic vectors of two operating days is input, and the aggregability of the residual load curve is output, which is set according to the start and stop conditions of the unit combination: if the start and stop conditions of the unit combination in each period of the two operating days are the same, the aggregability of the residual load curves of the two operating days is 1, otherwise it is 0. Train the SVM model, and the SVM finds the optimal hyperplane or function by adjusting its own parameters, so that the mapping relationship between input and output is as accurate as possible. The output value range of SVM is between (0, 1). The closer the output value is to 1, the more similar the residual load curves of the two operating days are and the stronger the aggregability is, and vice versa.

[0074] Step S220: Using the trained support vector machine, a residual load aggregatability matrix is ​​generated based on the residual load difference matrix; using the trained SVM model, a 365×365 residual load aggregatability matrix is ​​generated based on the residual load feature difference matrix, where is the input; is the output of the SVM model, indicating the and The residual load aggregability matrix is ​​a symmetric matrix with diagonal elements of 1. This is because the residual load curve of each operating day must be completely aggregable with itself, so the diagonal elements are At the same time, and The aggregability of the first running day is similar to that of the and The aggregability of the running days is the same, that is, ; The basis for judging the polymerizability in this embodiment is: When and The residual load curves of the operating days can be aggregated. The threshold is set according to the actual situation and experience. It is used to convert the continuous value of aggregatability into a discrete judgment result to facilitate the subsequent clustering operation.

[0075] This embodiment uses the powerful learning ability of the support vector machine to successfully construct a residual load aggregability matrix by training with the residual load feature vector difference as input and the aggregability set based on the start-stop situation of the unit combination as output. This matrix not only clearly quantifies the degree of aggregability of the residual load curve between each operating day, but also simplifies the analysis by using its symmetry and the characteristics of diagonal elements being 1. At the same time, setting a reasonable threshold converts continuous aggregability values ​​into discrete judgments, which provides an intuitive and effective basis for the subsequent clustering of operating days, greatly improves the scientificity and accuracy of clustering, and helps to more accurately classify the operating days of the power system, thereby optimizing the scheduling strategy and resource allocation in the power system production simulation, and improving the overall efficiency and stability of the power system operation.

[0076] Step S300: Iteratively cluster the operating days based on the residual load aggregatability matrix to extract typical days; specifically, by continuously merging operating days with high aggregatability, the number of operating days is gradually reduced, and new typical days are generated at the same time, until the aggregatability between all operating days is lower than the set threshold, thereby completing the clustering process.

[0077] Step S300 specifically includes:

[0078] Step S310: Based on the aggregability of the operating days, select the target operating day from the residual load aggregability matrix for aggregation to obtain the residual load curve of the typical day; specifically, in order to effectively cluster the operating days, this embodiment needs to find the two operating days with the largest aggregability from the residual load aggregability matrix and merge them. By merging the residual load curves of these two operating days, a new residual load curve of the typical day can be obtained;

[0079] Select the two running days with the greatest aggregability and , that is, the residual load aggregatability matrix The rows and columns corresponding to the elements with the largest values ​​in the matrix are aggregated; after aggregation, the typical day is obtained. The residual load curve can be expressed as:

[0080]

[0081] In the formula, Indicates a typical day The residual load curve of Indicates the operating day The residual load curve of Indicates the operating day The residual load curve of

[0082] Step S320: Delete the target operating day from the remaining load aggregatability matrix and add the typical day; specifically, after the aggregation of the operating day is completed, the aggregatability matrix needs to be updated to reflect the new clustering situation; delete the remaining load aggregatability matrix Mid-run day and The rows and columns represented by the The rows and columns represented by, at this time, the dimension of the residual load aggregatability matrix is ​​reduced from 365×365 to 364×364;

[0083] typical day With operation day The aggregability can be expressed as:

[0084]

[0085] In the formula, Indicates a typical day With operation day Aggregability; Indicates the operating day and Aggregability between Indicates the operating day and Aggregability between

[0086] Repeat steps S310 and S320 until all elements in the residual load aggregability matrix are less than 0.5, thereby obtaining Typical Day , It represents the number of typical days finally obtained, which reflects the number of different categories into which the running days are divided after clustering.

[0087] This embodiment performs clustering iterations on the operating days based on the residual load aggregatability matrix, merges the operating days with high aggregatability, effectively reduces the number of operating days, and generates representative typical days. It not only reduces the complexity of the power system production simulation and the amount of calculation, but also ensures that the clustering results can accurately reflect the similarities between operating days by continuously updating the aggregatability matrix. The typical days finally obtained can represent the characteristics of operating days of different categories, providing a concise and effective data basis for the subsequent solution of the unit combination model, helping to improve the efficiency and accuracy of power system operation planning and scheduling, and optimizing the allocation of power resources.

[0088] Step S400: With the goal of minimizing the total operating cost, the unit combination model is solved according to the typical day residual load curve to obtain the typical day start and stop results; specifically, the production simulation performs power system planning and operation analysis by solving the unit combination model; the unit combination model takes the goal of minimizing the total system operation cost and solves the start and stop mode and power generation output of the units in each time period;

[0089] Step S400 specifically includes:

[0090] Step S410: According to the typical day residual compliance curve, an objective function is constructed with the goal of minimizing the unit power generation cost and startup cost, and the unit commitment model is adjusted and solved based on the preset constraints. The expression of the objective function of the unit commitment model is:

[0091]

[0092] In the formula, For a typical day Medium Unit In the period The power generation cost is In the period contribution Function of is the total number of units; For a typical day Medium Unit In the period Start-up costs; The number of time periods in a day, that is, a day is divided into 24 time periods for power system operation analysis;

[0093] The constraints of the unit commitment model include:

[0094] The load balance constraint is configured as the total output of thermal power units in the target period of a typical day is equal to the total system load in that period minus the wind power output. The expression is:

[0095]

[0096] Where: For a typical day Medium thermal power unit In the period The output is the actual electric power generated by the unit during this period; For a typical day The total system load represents the total electricity demand that the power system needs to meet during this period; For a typical day Medium wind power output.

[0097] The system positive reserve constraint is configured as the sum of the maximum output of the thermal power units is greater than or equal to the sum of the system net load and the system positive reserve capacity. The expression is:

[0098]

[0099] Where: For a typical day Medium thermal power unit In the period Start and stop status; For a typical day Medium thermal power unit In the period Maximum output; For a typical day Mid-session The system has positive reserve capacity, which is the additional power generation capacity reserved for emergencies.

[0100] The system negative reserve constraint is configured as the minimum total output of thermal power units is not greater than the system net load plus the system negative reserve capacity, and the expression is:

[0101]

[0102] Where: For a typical day Medium thermal power unit In the period The minimum output of For a typical day Mid-session The system negative reserve capacity is used to cope with possible excess power in the system.

[0103] The upper and lower limits of the unit output are configured so that the output of the thermal power unit is between the minimum technical output and the maximum technical output. The expression is:

[0104]

[0105] Where: For a typical day Medium thermal power unit In the period Minimum technical output; For a typical day Medium thermal power unit In the period The maximum technical output.

[0106] The unit ramp rate constraint is configured to increase or decrease the unit output within the preset rate range. The expression is:

[0107]

[0108] Where: For thermal power units Maximum climbing rate; For thermal power units The maximum downhill climbing rate.

[0109] The minimum continuous start and stop time constraint of the unit is configured to maintain the preset continuous operation or shutdown time after the unit is started or stopped. The expression is:

[0110]

[0111] Where: For a typical day Medium thermal power unit In the period The continuous power-on time; For a typical day Medium thermal power unit In the period The continuous downtime time; Indicates the minimum continuous startup time of the unit, that is, the shortest time the unit needs to run continuously after startup; Indicates the minimum continuous downtime of the unit, that is, the shortest time the unit needs to be shut down continuously after being shut down;

[0112] Based on the above objective function and constraints, optimization algorithms such as mixed integer programming algorithm and heuristic algorithm are used to solve the unit combination model. and power generation output , so that the objective function reaches the minimum value, and all constraints are satisfied at the same time, and finally the start-up and shutdown results of a typical day are obtained. This embodiment takes into account the power generation cost and startup cost of the unit, and optimizes and solves under multiple constraints such as load balance, spare capacity, upper and lower limits of unit output, ramp rate, and minimum continuous start-up and shutdown time. It can scientifically and reasonably determine the start-up and shutdown status and power generation output of each unit on a typical day, providing an accurate and effective decision-making basis for the planning and operation of the power system.

[0113] Step S500: Based on the start-up and shutdown results of the typical day, the unit combination model is solved again according to the residual load curve of the operating day to obtain the output results of each period of the operating day; specifically, based on the typical day in step S400 Obtained As the unit start and stop results of the operating day represented by the typical day; for each operating day , if aggregated to a typical day Then the running day Medium thermal power unit In the period The start / stop state is set to ; For each operating day, based on the residual load curve of the operating day, combined with the objective function and constraints of the unit combination model used in step S400, the unit combination model of the operating day is constructed; the unit combination model of each operating day is solved using optimization algorithms such as mixed integer programming algorithms and heuristic algorithms. In the solution process, since the start and stop status of the unit has been determined based on the results of the typical day, the number of integer variables in the model is reduced, and the difficulty of solving the problem is relatively reduced. By solving the model, the output of the thermal power unit in each time period on each operating day can be obtained. , these output results are the final results of running the simulation.

[0114] This embodiment significantly reduces the complexity of solving the unit combination model by applying the start-up and shutdown results of typical days to the operating days, and solving the unit combination model again in combination with the actual remaining load curve of each operating day. Since the number of integer variables in the model is reduced, the difficulty of solving is greatly reduced and the calculation efficiency is improved. At the same time, the actual load characteristics of each operating day are fully considered, and more detailed and accurate unit output conditions for each operating day and each time period can be provided for the production simulation of the power system. It helps the power system to carry out more refined scheduling and planning, and improve the economy and reliability of the power system operation.

[0115] See also Figure 2 The technical solution of the second aspect of the present invention provides a power system production simulation system based on a support vector machine, using the power system production simulation method based on a support vector machine described in the technical solution of the first aspect of the present invention, the system comprises:

[0116] A data acquisition module configured to acquire a load forecast curve and a photovoltaic output forecast curve on an operation day;

[0117] a feature construction module configured to construct a residual load feature vector and a difference matrix according to a load forecast curve and a photovoltaic output forecast curve;

[0118] an aggregatability analysis module configured to construct a residual load aggregatability matrix for each operating day using a support vector machine based on a residual load characteristic difference matrix;

[0119] A clustering module configured to iteratively cluster the operating days based on the residual load aggregatability matrix and extract typical days;

[0120] The first solution module is configured to solve the unit commitment model according to the typical day residual load curve with the goal of minimizing the total operating cost, and obtain the typical day start-up and shutdown results;

[0121] The second solving module is configured to solve the unit combination model again based on the typical day start-up and shutdown results and the residual load curve of the operating day to obtain the output results of each period of the operating day.

[0122] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A power system production simulation method based on support vector machine, characterized in that: The method comprises: Obtain the load forecast curve and photovoltaic output forecast curve for the operating day, and construct the residual load feature vector and difference matrix, including: According to the load forecast curve and photovoltaic output curve on the operating day, the load characteristic points and photovoltaic output characteristic points are extracted; Construct a residual load characteristic vector based on the load characteristic points and the photovoltaic output characteristic points; Constructing a residual load difference matrix based on the residual load eigenvector; Based on the residual load characteristic difference matrix, the residual load aggregability matrix of each operating day is constructed using support vector machine, including: The difference of residual load feature vectors between operating days is used as input, and the aggregability of residual load curve is used as output to train the support vector machine. If the start and stop conditions of the unit combinations in each period of two operating days are the same, the aggregability of the residual load curve is 1. Using the trained support vector machine, the residual load aggregability matrix is ​​generated based on the residual load difference matrix; Based on the residual load aggregatability matrix, the operation days are clustered iteratively to extract typical days; With the goal of minimizing the total operating cost, the unit commitment model is solved according to the typical day residual load curve to obtain the typical day start-up and shutdown results; Based on the start-up and shutdown results of typical days, the unit combination model is solved again according to the residual load curve of the operating day to obtain the output results of each period of the operating day.

2. The power system production simulation method based on support vector machine according to claim 1, characterized in that: Based on the residual load aggregatability matrix, the operation days are clustered iteratively, and typical days are extracted, including: Based on the aggregability of the operating day, the target operating day is selected from the residual load aggregability matrix for aggregation to obtain the typical day residual load curve; Remove target operating days from the residual load aggregatability matrix and add typical days.

3. The power system production simulation method based on support vector machine according to claim 2, characterized in that: Based on the aggregability of the operating day, the target operating day is selected from the residual load aggregability matrix for aggregation to obtain the typical day residual load curve, including: The two operating days with the largest aggregability are selected for aggregation. After aggregation, the residual load curve of the typical day is obtained, which can be expressed as: In the formula, Indicates a typical day The residual load curve of Indicates the operating day The residual load curve of Indicates the operating day The residual load curve.

4. The power system production simulation method based on support vector machine according to any one of claims 1 to 3, characterized in that: With the goal of minimizing the total operating cost, the unit commitment model is solved according to the typical day residual load curve to obtain the typical day start-up and shutdown results, including: According to the typical day residual compliance curve, the objective function is constructed with the goal of minimizing the unit power generation cost and startup cost, and the unit combination model is solved based on the preset constraints.

5. The power system production simulation method based on support vector machine according to claim 4, characterized in that: The expression of the objective function is: In the formula, For a typical day Medium Unit In the period The power generation cost is In the period contribution Function of is the total number of units; For a typical day Medium Unit In the period Start-up costs; The number of periods in a day.

6. The power system production simulation method based on support vector machine according to claim 5, characterized in that: The preset constraints include: The load balance constraint is configured as follows: the total output of thermal power units during the target period of a typical day is equal to the total system load during that period minus the wind power output; The system positive reserve constraint is configured such that the sum of the maximum outputs of the thermal power units is greater than or equal to the sum of the system net load and the system positive reserve capacity; The system negative reserve constraint is configured so that the minimum total output of thermal power units is no greater than the system net load plus the system negative reserve capacity; The upper and lower limits of the unit output are configured so that the output of the thermal power unit is between the minimum technical output and the maximum technical output; The unit ramp rate constraint is configured to ensure that the unit output increases or decreases within a preset rate range; The minimum continuous start and stop time constraint of the unit is configured to maintain the preset continuous operation or shutdown time after the unit is started or shut down.

7. The power system production simulation method based on support vector machine according to claim 6, characterized in that: Based on the typical day start-up and shutdown results, the unit combination model is solved again according to the residual load curve of the operating day to obtain the output results of each period of the operating day, including: Based on the correspondence between typical days and operating days, the start and stop status of thermal power units in each period of typical days is applied to the corresponding operating days; According to the residual load curve of the operating day, the unit combination model is solved again in combination with the objective function and the preset constraints to obtain the output results of the thermal power units at different times of each operating day.

8. The power system production simulation system based on support vector machine is characterized by: The power system production simulation method based on support vector machine according to any one of claims 1 to 7 is adopted, and the system comprises: A data acquisition module configured to acquire a load forecast curve and a photovoltaic output forecast curve on an operation day; a feature construction module configured to construct a residual load feature vector and a difference matrix according to a load forecast curve and a photovoltaic output forecast curve; an aggregatability analysis module configured to construct a residual load aggregatability matrix for each operating day using a support vector machine based on a residual load characteristic difference matrix; A clustering module configured to iteratively cluster the operating days based on the residual load aggregatability matrix and extract typical days; The first solving module is configured to solve the unit commitment model according to the typical day residual load curve with the goal of minimizing the total operating cost, and obtain the typical day start-up and shutdown results; The second solving module is configured to solve the unit combination model again based on the typical day start-up and shutdown results and the residual load curve of the operating day to obtain the output results of each period of the operating day.

Images