Multi-source self-adaptive power generation coordination control method

Abstract

The invention relates to a multi-source self-adaptive power generation coordination control method. The method comprises the following steps: acquiring state data of each power generation device in a power grid; performing centralized training on the state data of each power generation device by adopting a federated learning mode to obtain a global power generation power prediction model; combining the obtained global power generation power prediction model with a corresponding power generation equipment physical model to obtain an overall prediction model; and dynamically adjusting a power generation control strategy and parameters of the overall prediction model according to the real-time state data and the optimization target of each power generation device, and obtaining a control strategy of each power generation device by adopting the adjusted overall prediction model. Compared with the prior art, the method has the advantages that the power generation power is accurately predicted and optimized, the optimal selection and dynamic adjustment of the power generation strategy are further realized, and the improvement of the system performance and efficiency is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of power generation control, and in particular to a multi-source adaptive power generation coordinated control method. Background Art

[0002] The automatic power generation control system currently on the market has the following defects:

[0003] 1. Single power source control: Traditional AGC systems mainly target a single type of power source and cannot fully utilize the advantages of multiple power sources, limiting the efficiency and flexibility of power generation coordination.

[0004] 2. Traditional control strategy: Traditional AGC systems use control strategies based on fixed rules and experience, lacking adaptability and sensitivity to real-time changes in the system, and cannot effectively deal with complex dynamic balance issues of power supply and demand. The load characteristics of ferroalloy production are special, with multiple load constraints such as ramping, stepping, and shutdown insulation. Traditional control strategies are difficult to deal with the changeable and complex dynamic balance issues of power supply and demand in ferroalloy production, which may cause energy waste or a decrease in ferroalloy production efficiency.

[0005] 3. Complexity of data processing: Traditional systems require a lot of complex data processing and calculations, resulting in slow system response and high operating costs. Summary of the invention

[0006] The purpose of the present invention is to provide a multi-source adaptive power generation coordinated control method in order to overcome the defects of the above-mentioned prior art that power generation control is based on fixed rules and experience-based control strategies and cannot effectively deal with the complex dynamic balance problem of power supply and demand.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A multi-source adaptive power generation coordinated control method comprises the following steps:

[0009] Obtain status data of each power generation equipment in the power grid;

[0010] The state data of each power generation equipment is centrally trained by using federated learning to obtain a global power generation prediction model; the global power generation prediction model is combined with the corresponding physical model of the power generation equipment to obtain an overall prediction model;

[0011] According to the real-time status data and optimization objectives of each power generation device, the power generation control strategy and parameters of the overall prediction model are dynamically adjusted, and the control strategy of each power generation device is obtained using the adjusted overall prediction model.

[0012] Furthermore, the data processing process of the overall prediction model is specifically as follows:

[0013] The power generation prediction model and the corresponding physical model of the power generation equipment are used to make rolling predictions on the power generation at future moments. A prediction feature parameter matrix is ​​constructed according to the input parameters of the power generation prediction model and the physical model of the power generation equipment. Based on the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, decision fusion and feature fusion are performed to obtain the mathematical and physical fusion results of the power generation capacity of each power generation equipment.

[0014] Furthermore, a granular feature alignment algorithm based on tensor product is used to process input parameters of the power generation prediction model and the physical model of the power generation equipment to obtain the prediction feature parameter matrix;

[0015] The tensor product-based granularity feature alignment algorithm comprises the following steps:

[0016] From the input parameters of the power generation prediction model and the physical model of the power generation equipment, select the features with large information content and perform feature extraction;

[0017] Align the input parameters after feature extraction and other input parameters of the power generation prediction model and the physical model of the power generation equipment in the time dimension;

[0018] Smoothing the aligned input parameters;

[0019] The output parameters after smoothing are normalized to obtain the prediction feature parameter matrix.

[0020] Furthermore, an improved non-dominated sorting genetic algorithm is used to perform the decision fusion and feature fusion, and the processing process of the improved non-dominated sorting genetic algorithm includes:

[0021] Based on the feature input in the prediction feature parameter matrix and the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, the feature and decision evaluation function are constructed, and the non-dominated sorting genetic algorithm is used to optimize the decision fusion coefficients in the feature and decision evaluation function, and finally the optimal mathematical and physical fusion result is obtained.

[0022] Furthermore, the calculation expressions of the features and decision evaluation functions are:

[0023]

[0024] In the formula, F(k1,k2,x i ) are the feature and decision evaluation function values, z i is the corresponding actual result, k1 is the first decision fusion coefficient, k2 is the second decision fusion coefficient, For feature x iThe characteristic input of the power generation prediction model after superimposing the random variable θ1, For feature x i The characteristic input of the physical model of the power generation equipment after superimposing the random variable θ2, y1 is the prediction result of the power generation prediction model, and y2 is the prediction result of the physical model of the power generation equipment.

[0025] Furthermore, the non-dominated sorting genetic algorithm comprises the following steps:

[0026] Generation of the initial population: The first decision fusion coefficient k1 and the second decision fusion coefficient k2 are used as two fragments, respectively, for mixed encoding to construct a particle structure, thereby generating the initial population according to the random principle;

[0027] Fast non-dominated sorting: stratify the population according to the non-inferior solution level of individual particles;

[0028] Virtual fitness: Sort the individuals in the same layer according to the corresponding features and decision evaluation function values, and calculate the corresponding crowding distance;

[0029] Selection operation: select individual particles based on the stratification results and crowding distance calculation results;

[0030] Crossover and mutation: Use the crossover operator to perform crossover operations on parent individuals;

[0031] Elite strategy: The elite strategy is used to retain the parent individuals with high fitness and directly assign them to the offspring.

[0032] Furthermore, the input parameters of the power generation prediction model and the power generation equipment physical model include the three-phase active power P of the ore-fired furnace. LF_abc (t), three-phase reactive power Q of the ore-arc furnace LF_abc (t), three-dimensional operating temperature of the submerged arc furnace T LF (t), total mass of molten ore in the submerged arc furnace m LF , load power demand curve P Load (t), impedance matrix RLC of the line in the system line , the loop matrix A of the network topology, the three-phase current I of each connecting line in the system line_abc (t), three-phase voltage U of each node of the system node_abc (t), meteorological data matrix M wind_solar (t), current wind speed v wind (t), current light l solar (t), rated wind speed v wind_rate , Maximum power generation wind speed v wind_max , minimum wind speed for power generation v wind_min , Rated light solar_rate ;

[0033] The output data of the power generation prediction model and the power generation equipment physical model include a wind power prediction curve P Wind_forecast (t), photovoltaic prediction curve P PV_forecast (t) and the electric power generation curve of the waste gas of the submerged arc furnace P gas_forecast (t).

[0034] Furthermore, the processing process of the overall prediction model also includes: according to the prediction results of the power generation capacity of each power generation equipment, combined with the corresponding load demand, an adjustable power generation control strategy is adopted to obtain the control strategy of each power generation equipment.

[0035] Furthermore, the power generation control strategy includes adjusting the optimization parameters of each power generation equipment within the corresponding constraints to achieve the best optimization target; the dynamic adjustment objects of the power generation control strategy include: the weights of each optimization target, the optimization parameters and the constraints.

[0036] Furthermore, the method is used for the control and management of new energy power generation and waste gas recovery power generation equipment in the ferroalloy industry.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] (1) The present invention combines the concepts of federated learning and data-physics fusion. In the algorithm, the power generation at future moments is predicted by the physical model and the data model respectively: in the data model, the data distributed on different power generation equipment are trained in a distributed manner by means of federated learning; in the physical model, a dynamic model of the ferroalloy multi-source power generation unit is established to obtain the prediction results of the physical model of the predicted power generation of each unit;

[0039] Finally, in order to achieve the fusion of prediction results, the input features are first aligned at a granular level. Then, the processed input features and the prediction results of the physical model and the data model are combined to increase the corresponding decision fusion coefficient. The feature and decision evaluation function are constructed in the form of minimum mean square error, and the non-dominated sorting genetic algorithm is used to determine the optimal decision fusion coefficient. The optimal fusion prediction result is obtained by fusion, which realizes the accurate prediction and optimization of power generation.

[0040] (2) The present invention adopts an adaptive optimization strategy to dynamically adjust the power generation control strategy and parameters according to real-time monitoring data and optimization objectives to achieve a balance between system frequency and net exchange power, and adopts an iterative optimization algorithm to improve system performance and efficiency.

[0041] (3) The present invention solves the problem of multi-source-load coordinated power generation in ferroalloy production: Since waste power generation equipment is commonly used in ferroalloy production, and the ferroalloy production load has variable and complex adjustment characteristics, the present invention innovatively considers the ferroalloy production source-load coupling model in the power generation coordination control system, determines the specific key input parameters for power generation prediction in the ferroalloy production load, solves the problem of multi-source-load coordinated power generation control in ferroalloy production, and effectively improves energy utilization or production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic flow chart of a multi-source adaptive power generation coordinated control method provided in an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of a data fusion process of a multi-source adaptive power generation coordinated control method provided in an embodiment of the present invention;

[0044] Figure 3 The present invention provides a schematic diagram of an adaptive optimization process of a multi-source adaptive power generation coordinated control method. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0046] 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 invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0048] In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are the directions or positional relationships in which the inventive product is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0049] It should be noted that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0050] In addition, the terms "horizontal", "vertical" and the like do not mean that the components are required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0051] Example 1

[0052] like Figure 1 As shown, this embodiment provides a multi-source adaptive power generation coordinated control method, comprising the following steps:

[0053] S1: Obtain status data of each power generation equipment in the power grid;

[0054] S2: Use federated learning to centrally train the status data of each power generation equipment to obtain a global power generation prediction model; combine the obtained global power generation prediction model with the corresponding power generation equipment physical model to obtain an overall prediction model;

[0055] S3: According to the real-time status data and optimization objectives of each power generation equipment, the power generation control strategy and parameters of the overall prediction model are dynamically adjusted, and the control strategy of each power generation equipment is obtained using the adjusted overall prediction model.

[0056] Preferably, the method is used for the control and management of new energy power generation and waste gas recovery power generation equipment in the ferroalloy industry to optimize the automatic generation control (AGC) function therein and realize automatic coordinated control of multiple types of power sources.

[0057] Specifically, Figure 2 As shown in Figure 1, the data processing process of the overall prediction model is as follows:

[0058] The power generation prediction model and the corresponding physical model of the power generation equipment are used to make rolling predictions on the power generation at future moments. A prediction feature parameter matrix is ​​constructed according to the input parameters of the power generation prediction model and the physical model of the power generation equipment. Based on the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, decision fusion and feature fusion are performed to obtain the mathematical and physical fusion results of the power generation capacity of each power generation equipment.

[0059] In addition, the processing of the overall prediction model also includes: according to the prediction results of the power generation capacity of each power generation equipment, combined with the corresponding load demand, an adjustable power generation control strategy is adopted to obtain the control strategy of each power generation equipment.

[0060] Specifically, the prediction feature parameter matrix is ​​obtained by processing the input parameters of the power generation prediction model and the physical model of the power generation equipment using a granular feature alignment algorithm based on tensor product;

[0061] The granular feature alignment algorithm based on tensor product includes the following steps:

[0062] From the input parameters of the power generation prediction model and the physical model of the power generation equipment, select the features with large information content and perform feature extraction;

[0063] Align the input parameters after feature extraction and other input parameters of the power generation prediction model and the physical model of the power generation equipment in the time dimension;

[0064] Smoothing the aligned input parameters;

[0065] The smoothed output parameters are normalized to obtain a prediction feature parameter matrix.

[0066] In this embodiment, the input parameters of the power generation prediction model and the power generation equipment physical model include: the three-phase active power P of the ore-fired furnace LF_abc (t), three-phase reactive power Q of the ore-arc furnace LF_abc (t), three-dimensional operating temperature of the submerged arc furnace T LF (t), total mass of molten ore in the submerged arc furnace m LF ; Load power demand curve P Load (t); impedance matrix RLC of the line in the system line , the loop matrix A of the network topology, the three-phase current I of each link line in the system line_abc (t), the three-phase voltage U of each node of the system node_abc (t); Meteorological data matrix M wind_solar (t), current wind speed v wind (t), current light l solar (t), rated wind speed v wind_rate , Maximum power generation wind speed v wind_max, minimum wind speed for power generation v wind_min , Rated light solar_rate .

[0067] The output parameters of the power generation prediction model and the power generation equipment physical model include: wind power prediction curve P Wind_forecast (t), photovoltaic prediction curve P PV_forecast (t), electric power generation curve of waste gas from submerged arc furnace P gas_forecast (t).

[0068] Assume that the parameter matrix of the input mathematical and physical fusion prediction is a cell array D = {x1,…,x k ,…,x K}, firstly, we need to use the granular feature alignment algorithm based on tensor product to process D:

[0069] 1) Three-dimensional operating temperature T of the submerged arc furnace LF (t) and other parameters with large information volume that is difficult to process. The SVM method is used to calculate the feature evaluation function and extract features:

[0070]

[0071] Among them, a is the total number of features; WA is the accuracy weight; SVM acuracy is the SVM accuracy; WF is the weight of the product and inverse of the selected features and penalty parameters.

[0072] 2) Align the input parameter time dimension:

[0073] x k =kron(x k ,B)

[0074] m=LCM(length(x1)…,length(x K )) / length(x k )

[0075] Where B is an m*1 ones matrix, kron represents the tensor product, and length represents the vector dimension.

[0076] 3) Smooth the aligned tensor parameter matrix:

[0077] X k (i,j)=(i-1)(X k (1,j+1)-X k (1,j)) / m+X k (1,j)

[0078] 4) Reconstruct parameter matrix x k , get the aligned array D′ and normalize it:

[0079] x k ′=(X k (:,1);…;X k (:,m))

[0080] D′={x1′,…,x k ′,…,x K ′}

[0081] D norm ′(i,j)=(D′(i,j)-min(D′(:,j))) / (max(D′(:,j))-min(D′(:,j)))

[0082] Preferably, an improved non-dominated sorting genetic algorithm is used for decision fusion and feature fusion, and the processing process of the improved non-dominated sorting genetic algorithm includes:

[0083] Based on the feature input in the prediction feature parameter matrix and the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, the feature and decision evaluation function are constructed, and the non-dominated sorting genetic algorithm is used to optimize the decision fusion coefficients in the feature and decision evaluation function, and finally the optimal mathematical and physical fusion result is obtained.

[0084] First, the established features and decision evaluation functions use the minimum mean square error model, and the corresponding calculation expression is:

[0085]

[0086] In the formula, F(k1,k2,x i ) are the feature and decision evaluation function values, z i is the corresponding actual result, k1 is the first decision fusion coefficient, k2 is the second decision fusion coefficient, For feature x i The characteristic input of the power generation prediction model after superimposing the random variable θ1, For feature x i The characteristic input of the physical model of the power generation equipment after superimposing the random variable θ2, y1 is the prediction result of the power generation prediction model, and y2 is the prediction result of the physical model of the power generation equipment.

[0087] Then the non-dominated sorting genetic algorithm Ad-NSGA-II is used for calculation. The main steps are as follows:

[0088] 1) Generation of the initial population. Particles use hybrid coding, fragment 1 is the first decision fusion coefficient k1, and fragment 2 is the second decision fusion coefficient k2.

[0089] The particle structure can be expressed as {k1(t1),k1(t2),…k1(t end ),k2(t1),k2(t2),…k2(t end )}. The initial population is generated according to the random principle, where the jth position of the ith particle is generated as follows:

[0090]

[0091] 2) Fast non-dominated sorting. Before the selection operation, the population is stratified according to the level of non-inferior solutions of the individuals. This algorithm needs to calculate two parameters n for each individual k in the population P. k and S k , where n k is the number of individuals that dominate individual k in the population, S k is the set of individuals dominated by individual k in the population. When performing fast non-inferiority stratification, you must first set i=1 and then enter the loop:

[0092] ① Find all n in the population l = 0, save l in the current set F i middle;

[0093] ②For the current set F i For each individual j in the j , traverse S j For each individual l, perform n l =n l -1;

[0094] ③Execute i=i+1 and return to ①.

[0095] Through the above loop operation, the entire population is stratified into F = (F1, F2, F3...).

[0096] 3) Virtual fitness. In order to maintain the diversity of individuals and prevent individuals from piling up locally, the NSGA-II algorithm first proposed the concept of virtual fitness. The specific calculation steps are:

[0097] ① Initialize the distance between individuals in the same layer. Let L[i] dis =0.

[0098] ② Arrange the individuals in the same layer in ascending order according to the mth objective function value. Let L = sort(L,m).

[0099] ③ There are n individuals in this layer, and the individuals on the edge of the layer are given a large number as the crowding distance L[1] dis =L[n] dis =∞.

[0100] ④ The individual in the middle of the ranking has a crowding distance of L[i]dis =L[i] dis +L[i+1].mL[i-1].m, where L[i].m is the mth objective function value of the ith individual.

[0101] ⑤For different objective functions, repeat steps ②-④.

[0102] 4) Selection operation. The selection operator is to avoid the loss of effective genes and make high-performance individuals survive with a greater probability, thereby improving global convergence and computational efficiency. After stratification and crowding distance calculation, each individual i in the population obtains two attributes: non-dominated order i rank and crowding distance L[i] dis . Using the contrast operator < n Double league selection for individuals, < n The specific execution process of the comparison operator is:

[0103] if(i rank <j rank )or[(i rank =j rank )and(L[i] dis <L[j] dis )]

[0104] Then(i>j)orElse(i <j)

[0105] < n The meaning of the selection is that if the non-dominated rankings of two individuals are different, the individual with the lower sequence number is selected; if two individuals are at the same level, the individual with the less crowded surroundings is selected.

[0106] 5) Crossover and mutation. The combination of crossover and mutation enables the genetic algorithm to have good local and global search performance. The crossover operator is used to perform crossover operations on the real-number-coded parent individuals, that is, for a given random crossover point, the parts of the two parent individuals on both sides of the crossover point are exchanged. In this paper, each bit of the offspring C1 and C2 is obtained by crossovering the corresponding bits of the parent generations P1 and P2 through the SBX (simulated binary crossover) operator. Take the jth bit as an example:

[0107]

[0108] Where u (j) is a random number uniformly distributed between [0,1]; η n is the crossover distribution index, that is, if the population size is N, then each generation executes η n N crossovers.

[0109] In this paper, each bit of the offspring C3 is obtained by mutating the corresponding bit of the parent P3 through the PM (polynomial mutation) operator. Taking the jth bit as an example, the PM operator execution strategy is as follows:

[0110] C 3(j) =P 3(j) +Δ (j)

[0111]

[0112] Where r (j) is a random number uniformly distributed between [0,1]; η m is the variation distribution index.

[0113] 6) Elite strategy. The elite strategy is to retain the excellent individuals in the parent generation and directly enter the offspring. It is a necessary condition for the genetic algorithm to converge with probability 1. The method used is:

[0114] ① The parent P t and offspring Q t Combined into a unified population: R t =P t ∪Q t , R t The number of individuals is 2N.

[0115] ② The population R t Quickly sort the non-dominated population and calculate the crowding distance of each individual. Select individuals one by one according to the level of the non-dominated hierarchy and the crowding distance until the number of individuals reaches N, forming a new parent population P. t+1 .

[0116] ③ On this basis, a new round of selection, crossover and mutation begins to form a new offspring population Q t+1 .

[0117] By using the time and dimension granularity alignment algorithm in the data fusion algorithm, and then performing feature extraction and analysis, comprehensively considering the actual operating conditions, historical data, and system model parameters, combining feature fusion with decision fusion technology, we can obtain the fused predicted power generation capacity curve and realize the prediction and optimization of power generation.

[0118] Specifically, Figure 3 As shown, the power generation control strategy includes adjusting the optimization parameters of each power generation equipment within the corresponding constraints to achieve the best optimization goal; the dynamic adjustment objects of the power generation control strategy include: the weights of each optimization goal, the optimization parameters and the constraints.

[0119] That is, this scheme establishes a dynamic characteristic model of ferroalloy load in the power generation optimization strategy to deal with the source-load coupled power generation control problem in ferroalloy production. The adaptive optimization strategy takes into account multiple optimization objectives including alloy output, system operation stability, and load. Based on the system stability and economic requirements, the system's safety, stability and operation status are monitored in real time, and the optimization target weights, optimization parameters, and constraints are dynamically adjusted. At the same time, the user's real-time optimization and adjustment requirements can be adjusted according to the user's real-time optimization requirements to maximize the flexibility and adaptability of the multi-source adaptive coordinated power generation system and achieve the system's active power exchange requirements and frequency stability requirements.

[0120] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A multi-source adaptive power generation coordinated control method, characterized in that: The following steps are involved: Obtain status data of each power generation equipment in the power grid; The state data of each power generation equipment is centrally trained by using federated learning to obtain a global power generation prediction model; the global power generation prediction model is combined with the corresponding physical model of the power generation equipment to obtain an overall prediction model; According to the real-time status data and optimization objectives of each power generation device, the power generation control strategy and parameters of the overall prediction model are dynamically adjusted, and the control strategy of each power generation device is obtained using the adjusted overall prediction model.

2. A multi-source adaptive power generation coordinated control method according to claim 1, characterized in that: The data processing process of the overall prediction model is specifically as follows: The power generation prediction model and the corresponding physical model of the power generation equipment are used to make rolling predictions on the power generation at future moments. A prediction feature parameter matrix is ​​constructed according to the input parameters of the power generation prediction model and the physical model of the power generation equipment. Based on the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, decision fusion and feature fusion are performed to obtain the mathematical and physical fusion results of the power generation capacity of each power generation equipment.

3. A multi-source adaptive power generation coordinated control method according to claim 2, characterized in that: Using a granularity feature alignment algorithm based on tensor product to process input parameters of the power generation prediction model and the power generation equipment physical model, to obtain the prediction feature parameter matrix; The tensor product-based granularity feature alignment algorithm comprises the following steps: From the input parameters of the power generation prediction model and the physical model of the power generation equipment, select the features with large information content and perform feature extraction; Align the input parameters after feature extraction and other input parameters of the power generation prediction model and the physical model of the power generation equipment in the time dimension; Smoothing the aligned input parameters; The output parameters after smoothing are normalized to obtain the prediction feature parameter matrix.

4. A multi-source adaptive power generation coordinated control method according to claim 2, characterized in that: An improved non-dominated sorting genetic algorithm is used to perform the decision fusion and feature fusion. The processing process of the improved non-dominated sorting genetic algorithm includes: Based on the feature input in the prediction feature parameter matrix and the power generation prediction results of the power generation prediction model and the physical model of the power generation equipment, the feature and decision evaluation function are constructed, and the non-dominated sorting genetic algorithm is used to optimize the decision fusion coefficients in the feature and decision evaluation function, and finally the optimal mathematical and physical fusion result is obtained.

5. A multi-source adaptive power generation coordinated control method according to claim 4, characterized in that: The calculation expressions of the features and decision evaluation functions are: In the formula, F(k1,k2,x i ) are the feature and decision evaluation function values, z i is the corresponding actual result, k1 is the first decision fusion coefficient, k2 is the second decision fusion coefficient, For feature x i The characteristic input of the power generation prediction model after superimposing the random variable θ1, For feature x i The characteristic input of the physical model of the power generation equipment after superimposing the random variable θ2, y1 is the prediction result of the power generation prediction model, and y2 is the prediction result of the physical model of the power generation equipment.

6. A multi-source adaptive power generation coordinated control method according to claim 4, characterized in that: The non-dominated sorting genetic algorithm comprises the following steps: Generation of the initial population: The first decision fusion coefficient k1 and the second decision fusion coefficient k2 are used as two fragments, respectively, for mixed encoding to construct a particle structure, thereby generating the initial population according to the random principle; Fast non-dominated sorting: stratify the population according to the non-inferior solution level of individual particles; Virtual fitness: Sort the individuals in the same layer according to the corresponding features and decision evaluation function values, and calculate the corresponding crowding distance; Selection operation: select individual particles based on the stratification results and crowding distance calculation results; Crossover and mutation: Use the crossover operator to perform crossover operations on parent individuals; Elite strategy: The elite strategy is used to retain the parent individuals with high fitness and directly assign them to the offspring.

7. A multi-source adaptive power generation coordinated control method according to claim 2, characterized in that: The input parameters of the power generation prediction model and the power generation equipment physical model include the three-phase active power P of the ore-fired furnace LF_abc (t), three-phase reactive power Q of the ore-arc furnace LF_abc (t), three-dimensional operating temperature of the submerged arc furnace T LF (t), total mass of molten ore in the ore-arcing furnace m LF , load power demand curve P Load (t), impedance matrix RC of the lines in the system line , the loop matrix A of the network topology, the three-phase current I of each connecting line in the system line_abc (t), three-phase voltage U of each node of the system node_abc (t), meteorological data matrix M wind_solar (t), current wind speed v wind (t), current light l solar (t), rated wind speed v wind_rate , Maximum power generation wind speed v wind_max , minimum wind speed for power generation v wind_min , Rated light solar_rate ; The output data of the power generation prediction model and the power generation equipment physical model include a wind power prediction curve P wind_forecast (t), photovoltaic prediction curve P PV_forecast (t) and the electric power generation curve of the waste gas of the submerged arc furnace P gas_forecast (t).

8. A multi-source adaptive power generation coordinated control method according to claim 2, characterized in that: The processing process of the overall prediction model also includes: according to the prediction results of the power generation capacity of each power generation equipment, combined with the corresponding load demand, an adjustable power generation control strategy is adopted to obtain the control strategy of each power generation equipment.

9. A multi-source adaptive power generation coordinated control method according to claim 8, characterized in that: The power generation control strategy includes adjusting the optimization parameters of each power generation device within the corresponding constraints to achieve the best optimization target; The dynamic adjustment objects of the power generation control strategy include: weights of various optimization objectives, optimization parameters and constraints.

10. The multi-source adaptive power generation coordinated control method according to claim 1, characterized in that: The method is used for controlling and managing new energy power generation and waste gas recovery power generation equipment in the ferroalloy industry.

Images