Wind power plant in-station loss calculation and reactive power optimization control method
By using the LSTM model to predict reactive power requirements in wind farms and dynamically adjusting the operating status of reactive power equipment in combination with reinforcement learning algorithms, the problems of low loss calculation accuracy and poor real-time optimization control in the prior art are solved, and more efficient reactive power optimization control is achieved.
Patent Information
- Application Number
- CN202510094027.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-21
- Publication Date
- 2025-05-16
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art has shortcomings in the calculation accuracy and dynamic response speed of reactive equipment loss in wind farms, resulting in unsatisfactory optimization control effect.
The LSTM model is used to build a time series prediction model, predict reactive power demand, and convert reactive power equipment losses into state vectors of reinforcement learning models. The operation status of reactive power equipment is dynamically adjusted through deep reinforcement learning algorithms to achieve real-time reactive power optimization control.
Through high-precision reactive demand prediction and dynamic adjustment of the operating status of reactive equipment, the equipment loss and turnover frequency are significantly reduced, and the operating efficiency and equipment service life of the wind farm are improved.
Smart Images
Figure CN120016614A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of wind power generation, and in particular to a method for calculating losses within a wind farm and optimizing reactive power control. Background Art
[0002] With the transformation of the global energy structure and the rapid development of renewable energy, wind power, as an important component of clean energy, has increased its share in the power system year by year. However, the operating characteristics of wind farms are significantly different from those of traditional power generation systems, and the volatility and uncertainty of their reactive power put forward high requirements on the stability of the power grid and the quality of power. In order to improve the operating efficiency and grid-connected stability of wind farms, domestic and foreign researchers have conducted extensive research in the field of reactive power optimization control, focusing on exploring how to reduce the reactive equipment loss in wind farms while meeting reactive power demand. Common reactive power optimization methods include methods based on traditional optimization algorithms, methods based on artificial intelligence algorithms, and optimization control methods combined with prediction models. These methods enable wind farms to better adapt to the needs of the power grid by regulating the operating status of reactive equipment in wind farms. However, due to the time dynamics, complexity and uncertainty of the reactive power demand of wind farms, the existing technology still has certain deficiencies in loss calculation accuracy, dynamic response speed and optimization control effect.
[0003] First, in terms of the calculation of reactive equipment loss, the existing technology usually uses static models or simplified mathematical expressions to approximate the operating loss of the equipment. Although this method is simple to calculate, it cannot accurately reflect the actual loss characteristics of the equipment under dynamic operation, especially when the reactive demand of the wind farm fluctuates greatly, the loss calculation error may increase significantly, thus affecting the overall optimization effect. Secondly, in terms of reactive optimization control, traditional methods are mostly based on fixed rules or static optimization models, which are difficult to respond to the changes in reactive demand of wind farms in real time, resulting in system regulation lag or unsatisfactory optimization effect. At the same time, although some optimization methods based on artificial intelligence have certain learning capabilities, they have limited processing capabilities for complex time series data and cannot accurately capture the dynamic changes in reactive demand of wind farms. This deficiency may cause problems such as unstable equipment operation status, increased losses or frequent equipment switching in the reactive optimization scheme of wind farms during operation, further restricting the operating efficiency and economy of wind farms. Summary of the invention
[0004] In view of the above existing problems, the present invention is proposed.
[0005] Therefore, the present invention provides a method for calculating losses and optimizing reactive power control within a wind farm, which solves the problems of low loss calculation accuracy and poor real-time performance of reactive power optimization control in the prior art.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0007] In the first aspect, the present invention provides a method for calculating losses within a wind farm and optimizing reactive power control, which includes collecting wind power operation data and preprocessing the wind power operation data; constructing a time series prediction model based on an LSTM model; obtaining a reactive demand prediction value based on the preprocessed wind power operation data through a time series prediction model, and calculating reactive equipment losses; converting reactive equipment losses into a state vector of a reinforcement learning model; defining a control strategy for reactive equipment of the reinforcement learning model according to the state vector of the reinforcement learning model, and dynamically adjusting the operating state of the reactive equipment; calculating the adjusted reactive equipment losses according to the adjustment result of the operating state of the reactive equipment, and realizing reactive optimization control of the wind farm by feedback optimization.
[0008] As a preferred solution of the method for calculating losses and optimizing reactive power control in a wind farm according to the present invention, the collected wind power operation data includes the output power, wind speed, wind direction, main transformer low-voltage side voltage meteorological data and the operating status of reactive equipment of the wind turbine;
[0009] The preprocessing of wind power operation data includes outlier processing, missing value filling and denoising.
[0010] As a preferred solution of the method for calculating the loss in a wind farm and optimizing reactive power control according to the present invention, the time series prediction model is constructed based on the LSTM model, and the specific steps are as follows:
[0011] Based on the pre-processed historical wind power operation data, it is converted into time series data through the sliding window method;
[0012] The LSTM model is used as the basic model;
[0013] The input layer receives time series data;
[0014] The LSTM layer captures long-term dependencies by stacking multiple LSTM layers;
[0015] The fully connected layer maps the output hidden state of the LSTM layer to the target value;
[0016] The final prediction result of the output layer;
[0017] Finally, a time series prediction model is constructed.
[0018] As a preferred solution of the method for calculating the loss in a wind farm and optimizing reactive power control according to the present invention, the reactive power demand forecast value is obtained based on the pre-processed wind power operation data through a time series prediction model, and the reactive power equipment loss is calculated. The specific steps are as follows:
[0019] Based on the time series prediction model, the preprocessed wind power operation data is input, and the reactive power demand at future moments is predicted using the time series prediction model. The predicted value is output. The expression is:
[0020]
[0021] in, represents the reactive power demand forecast value at time point t, t represents the index variable at the time point, and f LSTM represents the time series forecasting model, X t represents the preprocessed wind power operation data at time point t, W represents the weight of the LSTM model, and b represents the bias of the LSTM model;
[0022] According to the reactive power demand forecast value, the reactive power equipment loss is calculated as follows:
[0023]
[0024] Among them, L t represents the loss of reactive equipment at time point t, N represents the number of reactive equipment, k i represents the linear loss coefficient of the equipment, i represents the index variable of the reactive equipment, P i,t represents the active power loss of the ith reactive device at time t, c i Expressed as the nonlinear loss coefficient of the i-th reactive device, Q i,t represents the reactive power output of the ith reactive power device at time t, V i,t represents the operating voltage of the i-th reactive device at time point t, and ∈ represents a small amount that prevents the denominator from being zero.
[0025] As a preferred solution of the method for calculating the loss in a wind farm and optimizing reactive power control according to the present invention, the reactive power equipment loss is converted into a state vector of a reinforcement learning model, and the specific steps are as follows:
[0026] According to the reactive demand forecast value, reactive equipment loss, and the real-time acquired reactive equipment operation status, it is converted into the state vector s of reinforcement learning t , the expression of the state vector is:
[0027]
[0028] Among them, s t represents the state vector of reinforcement learning at time point t, Q t represents the current reactive output set of reactive devices at time t, V t represents the current operating voltage set of reactive equipment at time point t, u t represents the switching state set of reactive equipment at time point t, Δ trepresents the reactive power demand balance deviation at time point t, Y t Represents the historical operation feature matrix of the equipment at time point t.
[0029] As a preferred solution of the method for calculating the loss in the wind farm and optimizing the reactive power control according to the present invention, the control strategy of the reactive power equipment of the reinforcement learning model is defined according to the state vector of the reinforcement learning model, and the operating state of the reactive power equipment is dynamically adjusted. The specific steps are as follows:
[0030] Based on the state vector s t , use the DQN algorithm to generate the control strategy of reactive equipment, the expression is:
[0031] a t =π(s t );
[0032] Among them, a t represents the control strategy of reactive equipment at time point t, and π represents the strategy function;
[0033] Control strategy of reactive equipment t Including reactive power control action instructions of SVG, reactive power control action instructions of capacitor group and reactive power control action instructions of reactor group;
[0034] Based on the control strategy of reactive equipment, the reactive equipment of the wind farm is updated, and the expression is:
[0035]
[0036] in, Indicates the SVG reactive output after the update at time point t+1, represents the reactive power output at time point tSVG, a S Indicates the reactive power control action instruction of SVG;
[0037] According to the capacitor control strategy of reactive equipment, the reactive compensation amount of the capacitor is updated, and the expression is:
[0038]
[0039] in, represents the reactive output of the capacitor after updating at time point t+1, represents the reactive power output of the capacitor bank at time t, Represents the reactive power of a single group of capacitors, a C Indicates the reactive power control action instruction of the capacitor bank;
[0040] According to the reactor control strategy of reactive equipment, the reactive compensation amount of the reactor is updated, and the expression is:
[0041]
[0042] in, represents the reactive output of the reactor after updating at time point t+1, represents the reactive output of the reactor group at time point t, Indicates the reactive power of a single group of reactors, a R Indicates the reactive power control action command of the reactor group.
[0043] As a preferred solution of the method for calculating the loss in a wind farm and optimizing reactive power control according to the present invention, the adjusted reactive power equipment loss is calculated according to the adjustment result of the reactive power equipment operation state, and the specific steps are as follows:
[0044] According to the adjustment result of the reactive equipment operating status, the adjusted reactive equipment loss is calculated, and the expression is:
[0045]
[0046] in, represents the reactive equipment loss at time point t+1 after adjustment, represents the number of switching on and off of capacitors and reactors at time point t+1, k1 represents the proportionality coefficient of reactive power output loss, and k2 represents the proportionality coefficient of equipment switch operation loss;
[0047] According to the adjustment result of the operating status of the reactive equipment, the adjusted line loss is calculated, and the expression is:
[0048]
[0049] in, represents the line loss at time point t+1 after adjustment, R l Indicates the line resistance value, U t represents the voltage of the transmission line at time point t, P t Represents the effective power output of the fan at time point t.
[0050] As a preferred solution of the method for calculating the loss in a wind farm and optimizing reactive power control according to the present invention, the feedback optimization realizes the reactive power optimization control of the wind farm, and the specific steps are as follows:
[0051] According to the adjustment result of the reactive equipment operating status, the adjusted voltage deviation on the low-voltage side of the main transformer is calculated. The expression is:
[0052]
[0053] Among them, ΔU t+1 It represents the voltage deviation of the low-voltage side of the main transformer at time point t+1 after adjustment, U t+1represents the grid voltage at time t+1, C e Represents equivalent capacitance, U t Indicates the target voltage value at time point t;
[0054] The reactive equipment loss, line loss and voltage deviation on the low-voltage side of the main transformer are used as real-time feedback data. Combined with the adjustment results of the reactive equipment operating status, the time series prediction model and reinforcement learning model are optimized to achieve feedback optimization of reactive power optimization control of wind farms.
[0055] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for calculating in-station losses and optimizing reactive power control of a wind farm as described in the first aspect of the present invention is implemented.
[0056] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for calculating in-station losses and optimizing reactive power control of a wind farm as described in the first aspect of the present invention is implemented.
[0057] The beneficial effects of the present invention are as follows: high-precision prediction of reactive power demand is achieved through the LSTM model, providing reliable data support for optimization control; at the same time, the reactive equipment loss and operating status are converted into the state vector of the reinforcement learning model, and the operating status of the reactive equipment is dynamically adjusted using the deep reinforcement learning algorithm to meet the real-time reactive power demand response; through the feedback optimization mechanism, the reinforcement learning model continuously optimizes the control strategy, effectively reduces equipment loss and switching frequency, and improves equipment operating efficiency and service life. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.
[0059] Figure 1 This is a flow chart of the method for calculating losses within a wind farm and optimizing reactive power control in Example 1.
[0060] Figure 2 This is a flowchart for constructing a time series prediction model in Example 1. DETAILED DESCRIPTION
[0061] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.
[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0064] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for calculating losses in a wind farm and optimizing reactive power control, comprising the following steps:
[0065] S1. Collect wind power operation data including wind turbine output power, wind speed, wind direction, main transformer low-voltage side voltage meteorological data and operating status of reactive equipment;
[0066] Preprocessing of wind power operation data includes outlier processing, missing value filling and denoising;
[0067] It should be noted that outlier processing ensures the authenticity of the data, common statistical methods or time series models are used to detect and correct outliers, missing value filling ensures the integrity of the data, appropriate methods are selected to fill missing values, denoising improves the smoothness and effectiveness of the data, and effective information is extracted through filtering or deep learning.
[0068] S2. Build a time series prediction model based on the LSTM model;
[0069] Based on the pre-processed historical wind power operation data, it is converted into time series data through the sliding window method;
[0070] It should be noted that the sliding window method is an efficient time series feature extraction method. The window size and step size can be adjusted according to needs to adapt to different time dimensions and prediction needs. It is not only suitable for wind power prediction, but also for other time series problems.
[0071] The LSTM model is used as the basic model;
[0072] The input layer receives time series data;
[0073] The LSTM layer captures long-term dependencies by stacking multiple LSTM layers;
[0074] The fully connected layer maps the output hidden state of the LSTM layer to the target value;
[0075] The final prediction result of the output layer;
[0076] Finally, a time series forecasting model is constructed;
[0077] It should be noted that through the LSTM model, the entire process realizes a complete closed loop from data collection to final optimization, ensuring the effective calculation of wind farm losses and reactive power optimization control. In particular, the application of time series prediction based on the LSTM model can more accurately predict future load changes, plan reactive power compensation measures in advance, and reduce uncertainty risks.
[0078] S3. Based on the pre-processed wind power operation data, the reactive power demand forecast value is obtained through the time series forecasting model, and the reactive power equipment loss is calculated;
[0079] Based on the time series prediction model, the preprocessed wind power operation data is input, and the reactive power demand at future moments is predicted using the time series prediction model. The predicted value is output. The expression is:
[0080]
[0081] in, represents the reactive power demand forecast value at time point t, t represents the index variable at the time point, and f LSTM represents the time series forecasting model, X t represents the preprocessed wind power operation data at time point t, W represents the weight of the LSTM model, and b represents the bias of the LSTM model;
[0082] It should be noted that by using the LSTM-based time series prediction model, a multi-layer LSTM network is trained on the pre-processed wind power operation data to capture long-term dependencies and accurately predict the reactive power demand at future times. The trained model generates reactive power demand forecast values by receiving the latest data in real time, guiding the intelligent dispatching to dynamically adjust the reactive power compensation strategy, thereby optimizing performance and reducing losses.
[0083] According to the reactive power demand forecast value, the reactive power equipment loss is calculated as follows:
[0084]
[0085] Among them, L t represents the loss of reactive equipment at time point t, N represents the number of reactive equipment, k i represents the linear loss coefficient of the equipment, i represents the index variable of the reactive equipment, P i,t represents the active power loss of the ith reactive device at time t, c iExpressed as the nonlinear loss coefficient of the i-th reactive device, Q i,t represents the reactive power output of the ith reactive power device at time t, V i,t represents the operating voltage of the i-th reactive device at time point t, ∈ represents a small amount to prevent the denominator from being zero;
[0086] It should be noted that the reactive equipment loss formula evaluates the total loss L at time point t by comprehensively calculating the linear loss and nonlinear loss of the equipment. t ; Among them, linear loss is proportional to the active loss of the equipment, and nonlinear loss is inversely proportional to the reactive output and the square of the operating voltage, while considering the stability of the denominator ∈; the formula is applicable to different types of reactive compensation equipment and can be used for real-time loss assessment, operation optimization and equipment maintenance, and by optimizing the reactive output Q i,t and operating voltage V i,t To minimize losses and improve grid operation efficiency and economy.
[0087] S4, converting reactive equipment loss into a state vector of a reinforcement learning model;
[0088] According to the reactive demand forecast value, reactive equipment loss, and the real-time acquired reactive equipment operation status, it is converted into the state vector s of reinforcement learning t , the expression of the state vector is:
[0089]
[0090] Among them, s t represents the state vector of reinforcement learning at time point t, Q t represents the current reactive output set of reactive devices at time t, V t represents the current operating voltage set of reactive equipment at time point t, u t represents the switching state set of reactive equipment at time point t, Δ t represents the reactive power demand balance deviation at time point t, Y t Represents the historical operation characteristic matrix of the equipment at time point t;
[0091] It should be noted that the state vector s t The specific sources and physical meanings of the variables in , especially the reactive power demand balance deviation Δ t The calculation method of the device historical operation characteristic matrix Y t The construction method is used to ensure that the state vector can fully and accurately reflect the current operating state of the system and provide effective information for reinforcement learning decision-making.
[0092] S5. According to the state vector of the reinforcement learning model, define the control strategy of the reactive equipment of the reinforcement learning model, and dynamically adjust the operating state of the reactive equipment;
[0093] Based on the state vector s t , use the DQN algorithm to generate the control strategy of reactive equipment, the expression is:
[0094] a t =π(s t );
[0095] Among them, a t represents the control strategy of reactive equipment at time point t, and π represents the strategy function;
[0096] It should be noted that the policy function π is a deep reinforcement learning model based on DQN, using the current state vector s t As input, predict the Q value of each possible action and select the optimal action a t as a regulatory strategy.
[0097] Control strategy of reactive equipment t Including reactive power control action instructions of SVG, reactive power control action instructions of capacitor group and reactive power control action instructions of reactor group;
[0098] It should be noted that the reactive power control action command of SVG, the reactive power control action command of capacitor bank and the reactive power control action command of reactor bank respectively represent a S 、a C and a R .
[0099] Based on the control strategy of reactive equipment, the reactive equipment of the wind farm is updated, and the expression is:
[0100]
[0101] in, Indicates the SVG reactive output after the update at time point t+1, represents the reactive power output at time point tSVG, a S Indicates the reactive power control action instruction of SVG;
[0102] It should be noted that a S It is a reactive power control action instruction generated based on the current state, which is used to adjust the reactive power output of SVG to meet the reactive power demand of wind farms, maintain voltage stability and optimize grid operation. It ensures the feasibility of the control result by combining the equipment operation restrictions, and realizes efficient and intelligent dynamic control through reinforcement learning algorithm.
[0103] According to the capacitor control strategy of reactive equipment, the reactive compensation amount of the capacitor is updated, and the expression is:
[0104]
[0105] in, represents the reactive output of the capacitor after updating at time point t+1, represents the reactive power output of the capacitor bank at time t, Represents the reactive power of a single group of capacitors, a C Indicates the reactive power control action instruction of the capacitor bank;
[0106] It should be noted that a C The number of capacitor banks to be switched is determined by adjusting the switching status of the capacitor banks dynamically, which can achieve real-time balance of reactive power demand, voltage regulation and optimization of operating costs. Combined with the control actions generated by the reinforcement learning algorithm, it can guide the operation and control of the capacitor banks efficiently and intelligently.
[0107] According to the reactor control strategy of reactive equipment, the reactive compensation amount of the reactor is updated, and the expression is:
[0108]
[0109] in, represents the reactive output of the reactor after updating at time point t+1, represents the reactive output of the reactor group at time point t, Indicates the reactive power of a single group of reactors, a R Indicates the reactive power control action instruction of the reactor group;
[0110] It should be noted that by dynamically adjusting the switching state of the reactor group, real-time balancing of reactive power demand, voltage regulation and optimization of operating costs can be achieved. Combined with the control actions generated by the reinforcement learning algorithm, the operation control of the reactor group can be guided efficiently and intelligently.
[0111] S6. Calculate the adjusted reactive equipment loss according to the adjustment result of the reactive equipment operating state;
[0112] According to the adjustment result of the reactive equipment operating status, the adjusted reactive equipment loss is calculated, and the expression is:
[0113]
[0114] in, represents the reactive equipment loss at time point t+1 after adjustment, represents the number of switching on and off of capacitors and reactors at time point t+1, k1 represents the proportionality coefficient of reactive power output loss, and k2 represents the proportionality coefficient of equipment switch operation loss;
[0115] It should be noted that the reactive equipment loss is divided into the operating loss of SVG and the switching loss of capacitors and reactors. By reasonably regulating the operating status of reactive equipment, the total equipment loss can be minimized while meeting the reactive demand, providing a quantitative basis for the efficient operation of reactive equipment and can be used to optimize the design and implementation of regulation strategies.
[0116] According to the adjustment result of the operating status of the reactive equipment, the adjusted line loss is calculated, and the expression is:
[0117]
[0118] in, represents the line loss at time point t+1 after adjustment, R l Indicates the line resistance value, U t represents the voltage of the transmission line at time point t, P t represents the effective power output of the fan at time t;
[0119] It should be noted that by adjusting the output of reactive equipment, the transmission of reactive power in the line is reduced, thereby reducing the line current. and line loss, improve U t It can effectively reduce line current, thereby reducing line losses, and provide reactive power compensation close to the load side, which can reduce the need for long-distance reactive power transmission and reduce losses.
[0120] S7, feedback optimization to achieve reactive power optimization control of wind farm;
[0121] According to the adjustment result of the reactive equipment operating status, the adjusted voltage deviation on the low-voltage side of the main transformer is calculated. The expression is:
[0122]
[0123] Among them, ΔU t+1 It represents the voltage deviation of the low-voltage side of the main transformer at time point t+1 after adjustment, U t+1 represents the grid voltage at time t+1, C e Represents equivalent capacitance, U t Indicates the target voltage value at time point t;
[0124] It should be noted that the influence of reactive equipment operation adjustment on the low-voltage side voltage of the main transformer is calculated, and the voltage deviation ΔU is calculated. t+1 Used to evaluate the regulation effect. By optimizing the output of reactive equipment, the voltage deviation can be minimized to ensure stable operation of the grid voltage while meeting the target voltage requirements.
[0125] The reactive equipment loss, line loss and voltage deviation on the low-voltage side of the main transformer are used as real-time feedback data. Combined with the adjustment results of the reactive equipment operating status, the time series prediction model and reinforcement learning model are optimized to achieve feedback optimization of reactive power optimization control of wind farms.
[0126] It should be noted that by dynamically sensing the operating status and adjusting the control strategy in real time, the losses can be minimized while meeting the voltage stability target, thereby improving the operating efficiency and economy of the wind farm.
[0127] This embodiment also provides a computer device, which is suitable for the case of wind farm in-station loss calculation and reactive power optimization control method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the wind farm in-station loss calculation and reactive power optimization control method proposed in the above embodiment.
[0128] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.
[0129] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for calculating the losses in a wind farm and optimizing the reactive power control proposed in the above embodiment is implemented. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a disk or an optical disk.
[0130] In summary, the present invention achieves high-precision prediction of reactive power demand through: LSTM model, providing reliable data support for optimization control; at the same time, the reactive equipment loss and operating status are converted into the state vector of the reinforcement learning model, and the deep reinforcement learning algorithm is used to dynamically adjust the operating status of the reactive equipment to meet the real-time reactive power demand response; through the feedback optimization mechanism, the reinforcement learning model continuously optimizes the control strategy, effectively reduces equipment loss and switching frequency, and improves equipment operating efficiency and service life.
[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A method for calculating losses and optimizing reactive power control in a wind farm, characterized in that: include, Collect wind power operation data and pre-process the wind power operation data; Based on the LSTM model, build a time series prediction model; Based on the pre-processed wind power operation data, the reactive power demand forecast value is obtained through the time series prediction model, and the reactive power equipment loss is calculated; Convert reactive equipment loss into the state vector of the reinforcement learning model; According to the state vector of the reinforcement learning model, the control strategy of the reactive equipment of the reinforcement learning model is defined to dynamically adjust the operating state of the reactive equipment; According to the adjustment results of the reactive equipment operating status, the adjusted reactive equipment loss is calculated, and feedback optimization is used to achieve reactive power optimization control of the wind farm.
2. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 1, characterized in that: The collected wind power operation data includes the output power, wind speed, wind direction of the wind turbine, the main transformer low-voltage side voltage meteorological data and the operating status of the reactive equipment; The preprocessing of wind power operation data includes outlier processing, missing value filling and denoising.
3. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 2, characterized in that: The time series prediction model is constructed based on the LSTM model. The specific steps are as follows: Based on the pre-processed historical wind power operation data, it is converted into time series data through the sliding window method; The LSTM model is used as the basic model; The input layer receives time series data; The LSTM layer captures long-term dependencies by stacking multiple LSTM layers; The fully connected layer maps the output hidden state of the LSTM layer to the target value; The final prediction result of the output layer; Finally, a time series prediction model is constructed.
4. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 3, characterized in that: Based on the pre-processed wind power operation data, the reactive power demand forecast value is obtained through the time series forecasting model, and the reactive power equipment loss is calculated. The specific steps are as follows: Based on the time series prediction model, the preprocessed wind power operation data is input, and the reactive power demand at future moments is predicted using the time series prediction model. The predicted value is output. The expression is: in, represents the reactive power demand forecast value at time point t, t represents the index variable at the time point, and f LSTM represents the time series forecasting model, X t represents the preprocessed wind power operation data at time point t, W represents the weight of the LSTM model, and b represents the bias of the LSTM model; According to the reactive power demand forecast value, the reactive power equipment loss is calculated as follows: Among them, L t represents the loss of reactive equipment at time point t, N represents the number of reactive equipment, k i represents the linear loss coefficient of the equipment, i represents the index variable of the reactive equipment, P i,t represents the active power loss of the ith reactive device at time t, c i Expressed as the nonlinear loss coefficient of the i-th reactive device, Q i,t represents the reactive power output of the ith reactive power device at time t, V i,t represents the operating voltage of the i-th reactive device at time point t, and ∈ represents a small amount that prevents the denominator from being zero.
5. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 4, characterized in that: The reactive equipment loss is converted into the state vector of the reinforcement learning model. The specific steps are as follows: According to the reactive demand forecast value, reactive equipment loss, and the real-time acquired reactive equipment operation status, it is converted into the state vector s of reinforcement learning t , the expression of the state vector is: Among them, s t represents the state vector of reinforcement learning at time point t, Q t Represents the current reactive output set of reactive devices at time t, V t represents the current operating voltage set of reactive equipment at time point t, u t represents the switching state set of reactive equipment at time point t, Δ t represents the reactive power demand balance deviation at time point t, Y t Represents the historical operation feature matrix of the equipment at time point t.
6. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 5, characterized in that: According to the state vector of the reinforcement learning model, the control strategy of the reactive equipment of the reinforcement learning model is defined to dynamically adjust the operating state of the reactive equipment. The specific steps are as follows: Based on the state vector s t , use the DQN algorithm to generate the control strategy of reactive equipment, the expression is: a t =π(s t ); Among them, a t represents the control strategy of reactive equipment at time point t, and π represents the strategy function; Control strategy of reactive equipment t Including reactive power control action instructions of SVG, reactive power control action instructions of capacitor group and reactive power control action instructions of reactor group; Based on the control strategy of reactive equipment, the reactive equipment of the wind farm is updated, and the expression is: in, Indicates the SVG reactive output after the update at time point t+1, represents the reactive power output at time point tSVG, a S Indicates the reactive power control action instruction of SVG; According to the capacitor control strategy of reactive equipment, the reactive compensation amount of the capacitor is updated, and the expression is: in, represents the reactive output of the capacitor after updating at time point t+1, represents the reactive power output of the capacitor bank at time t, Represents the reactive power of a single group of capacitors, a C Indicates the reactive power control action instruction of the capacitor bank; According to the reactor control strategy of reactive equipment, the reactive compensation amount of the reactor is updated, and the expression is: in, represents the reactive output of the reactor after updating at time point t+1, represents the reactive output of the reactor group at time point t, Indicates the reactive power of a single group of reactors, a R Indicates the reactive power control action command of the reactor group.
7. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 6, characterized in that: The reactive equipment loss after adjustment is calculated according to the adjustment result of the reactive equipment operation state. The specific steps are as follows: According to the adjustment result of the reactive equipment operating status, the adjusted reactive equipment loss is calculated, and the expression is: in, represents the reactive equipment loss at time point t+1 after adjustment, represents the number of switching on and off of capacitors and reactors at time point t+1, k1 represents the proportionality coefficient of reactive power output loss, and k2 represents the proportionality coefficient of equipment switch operation loss; According to the adjustment result of the operating status of the reactive equipment, the adjusted line loss is calculated, and the expression is: in, represents the line loss at time point t+1 after adjustment, R l Indicates the line resistance value, U t represents the voltage of the transmission line at time point t, P t Represents the effective power output of the fan at time point t.
8. The method for calculating losses and optimizing reactive power control in a wind farm according to claim 7, characterized in that: The feedback optimization realizes the reactive power optimization control of the wind farm, and the specific steps are as follows: According to the adjustment result of the reactive equipment operating status, the adjusted voltage deviation on the low-voltage side of the main transformer is calculated. The expression is: Among them, ΔU t+1 It represents the voltage deviation of the low-voltage side of the main transformer at time point t+1 after adjustment, U t+1 represents the grid voltage at time t+1, C e Represents equivalent capacitance, U t Indicates the target voltage value at time point t; The reactive equipment loss, line loss and voltage deviation on the low-voltage side of the main transformer are used as real-time feedback data. Combined with the adjustment results of the reactive equipment operating status, the time series prediction model and reinforcement learning model are optimized to achieve feedback optimization of reactive power optimization control of wind farms.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for calculating losses within a wind farm and optimizing reactive power control according to any one of claims 1 to 8 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for calculating wind farm in-station losses and optimizing reactive power control as claimed in any one of claims 1 to 8 are implemented.