Quick starting method for shortening starting operation time of static var generator

Abstract

The invention relates to a quick starting method for shortening the starting operation time of a static var generator. The static var generator is configured to be in a preset standby state in advance in an off-peak operation period; projects related to operation of the static var generator are divided into core items and non-core items; for the core item, when the core item is started, comprehensive detection is executed; for non-core items, fault rate evaluation is carried out, when an evaluation result is higher than a set fault rate threshold value, detection is carried out during starting, and when the evaluation result is lower than the set fault rate threshold value, detection is skipped during starting; setting output calculation parameters; the method has the advantages that in the idle period, key operation parameters of the SVG are preset, the optimal working state can be quickly achieved after starting, and therefore time is saved; when items related to SVG operation are detected, the items needing to be detected are divided into core items and non-core items, the core items must be detected, and the non-core items are partially detected and partially skipped, so that the starting time is greatly shortened.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, and particularly relates to a quick start method for shortening the start-up operation time of a static var generator. BACKGROUND

[0002] As the most important traditional festival in China, the Spring Festival usually brings about the phased adjustment of production activities and the significant change of residents' electricity consumption mode. During the period from 15 days before the Spring Festival to the 29th day of the lunar year, many industrial enterprises gradually reduce production, and some even stop work and go on holiday, resulting in the decrease of power grid load. However, within a week after the Spring Festival holiday, with the resumption of work and production of enterprises, the power grid load gradually rises again.

[0003] The static var generator (SVG), a key reactive power compensation device in the power system, has an important influence on the stability of the power grid in terms of its running state and response speed. However, the existing SVG device has the problem of long start-up operation time when it is restarted after shutdown during the Spring Festival, which seriously affects its running efficiency and response ability. Especially during the pre-festival and post-festival periods when the power grid load fluctuates frequently and the reactive power demand changes dramatically, the long start-up time may result in the SVG failing to be put into operation in time, causing the aggravation of power grid voltage fluctuation and the deterioration of power factor, and thus affecting the power supply quality and system stability.

[0004] Through actual investigation, it is found that the main reasons for the long SVG start-up time include:

[0005] 1. The state preparation and detection process is complex and time-consuming: Before starting, the traditional SVG needs to set multiple operating parameters, such as DC side voltage, output current reference value, temperature control threshold, etc. In addition to parameter setting, it also needs to comprehensively check various components closely related to SVG operation, including but not limited to electrical parameters, protection devices, cooling systems, ventilation equipment, etc. Through detection, it is confirmed whether each component is in good condition and whether there is a potential fault risk. The above process is crucial to ensure the safe and reliable operation of the SVG, but at the same time, it also increases the time cost of start-up preparation;

[0006] 2. The output calculation parameters depend on manual setting, and the configuration efficiency is low: The control parameters of the current SVG are mostly set manually by the operation and maintenance personnel according to experience, which is difficult to quickly adapt to the dynamic changes of power grid conditions, affecting the quick commissioning of the device;

[0007] 3. Lack of ability to predict the trend of load changes: The existing SVG control system fails to effectively combine the load prediction model for advance preparation and strategy adjustment, reducing the response ability of the device to load mutation.

[0008] Based on this, the present application is proposed. SUMMARY

[0009] The application aims to provide a quick start method for shortening the start-up operation time of a static var generator, thereby significantly shortening the start-up operation time of an SVG, improving the operation efficiency of the device and the power grid support capability.

[0010] In order to achieve the above-mentioned purpose, the technical scheme of the application is as follows:

[0011] A quick start method for shortening the start-up operation time of a static var generator, comprising the following steps:

[0012] S10. In a non-peak operation period, the static var generator is configured to a preset standby state in advance, and the preset standby state includes the setting of the operation parameters of the static var generator;

[0013] S20. Items related to the operation of the static var generator are divided into core items and non-core items; for the core items, comprehensive detection is performed at start-up; for the non-core items, a fault prediction model based on historical operation data and a machine learning algorithm is used to evaluate the fault rate of the non-core items at the end or during the last operation cycle, and when the evaluation result of a certain non-core item is higher than the set fault rate threshold, detection is performed at start-up, and when the evaluation result of a certain non-core item is lower than the set fault rate threshold, detection is skipped at start-up;

[0014] S30. Output calculation parameter setting;

[0015] S40. Static var generator start-up completion.

[0016] Further, the core items that need to be detected include but are not limited to whether the electrical parameters meet the operation requirements, whether the protection devices are normal, whether the temperature is in a normal state, and whether the physical connection between devices is correct and not loose;

[0017] The non-core items that need to be evaluated include but are not limited to whether the environmental conditions affect the start-up of the device, whether the redundant devices are normal, whether the cooling system is normal, and whether the ventilation device is normal.

[0018] Further, the fault prediction model uses a long short-term memory network (LSTM) or an XGBoost algorithm, and uses features in historical data as input to evaluate the fault probability of the non-core items.

[0019] Further, the output calculation parameter setting includes the following process: historical operation data is analyzed by a machine learning algorithm, the current power grid working condition is automatically matched, and an optimal output calculation parameter combination is dynamically generated.

[0020] Further, the machine learning model of the output calculation parameter is as follows:

[0021] P * (X)=WT X+b;

[0022]

[0023] Wherein, X represents an input feature vector, represents the related parameters of power grid working condition;W represents a weight vector, W T is the transpose of W;b represents a bias term, which is a constant value;P* represents the optimal output parameter setting value obtained by model prediction;n is the sample number;L(W,b) is the loss function;X i is the feature vector of the i th sample, Y i is the real value of the corresponding optimal output parameter.

[0024] Further, in the step S40, after the static var generator is started, the power grid load information is collected in real time, and the output power of the static var generator is dynamically adjusted in combination with a load change trend prediction model.

[0025] Further, the load change trend prediction model adopts an ARIMA(p,d,q) model.

[0026] Further, the failure rate threshold is dynamically adjusted by using a dynamic adjustment model, and the dynamic adjustment model is as follows:

[0027]

[0028] Wherein, θ represents the set failure rate threshold;MTBF is the average failure-free working time;R(t) represents the reliability evaluation function of a certain non-core item at the current time;α and β are weighting coefficients, which are artificially set.

[0029] The advantages of the present application are:

[0030] 1. In the idle period, the key operating parameters of the SVG are pre-set, such as the DC side voltage, the output current reference value, the temperature control threshold, etc., to ensure that the SVG can quickly reach the optimal working state after starting, thereby saving a part of time;

[0031] 2. When detecting the items related to the operation of the SVG, the items to be detected are divided into core items and non-core items, the core items are detected, and the non-core items are partially detected and partially skipped, which can greatly shorten the starting time, and for the items to be detected in the non-core items, a failure prediction model constructed by using a machine learning algorithm is used to judge whether the non-core items need to be detected, to ensure the safety and reliability of the equipment;

[0032] 3. The output calculation parameter setting is based on machine learning algorithm analysis of historical operation data, automatic matching of current power grid working condition, dynamic generation, reduction of manual intervention, rapid adaptation to dynamic changes of power grid working condition, and rapid commissioning of equipment;

[0033] 4. By monitoring and predicting the change of power grid load in real time, the output power of the SVG can be adjusted in advance to avoid system instability or low efficiency caused by sudden load change, and the stability of the power grid operation is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 Flowchart of the quick start method for shortening the start-up operation time of the static var generator in the embodiment. DETAILED DESCRIPTION

[0035] The present application will be further described in detail below in conjunction with the embodiments. It should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like in the text indicate the orientation or positional relationship shown in the coordinate system of the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0036] The embodiment proposes a quick start method for shortening the start-up operation time of the static var generator, as shown in Figure 1 The method comprises the following steps:

[0037] S10. In the off-peak operating period, the static var generator is configured to a preset standby state, and the preset standby state comprises the setting of the operating parameters of the static var generator;

[0038] S20. The items related to the operation of the static var generator are divided into core items and non-core items; for the core items, comprehensive detection is performed at start-up; for the non-core items, a fault prediction model based on historical operation data and machine learning algorithm is used to evaluate the fault rate of the non-core items at the end or during the last operation cycle, and when the evaluation result of a certain non-core item is higher than the set fault rate threshold, detection is performed at start-up, and when the evaluation result of a certain non-core item is lower than the set fault rate threshold, detection is skipped at start-up;

[0039] S30. Output calculation parameter setting;

[0040] S40. Static var generator start-up is completed.

[0041] In step S10, the operating parameters of the SVG are pre-set, including but not limited to voltage, current, temperature and other key parameters, so that when the power grid load suddenly changes or needs to be adjusted urgently, the SVG can quickly switch from standby state to working state, shortening the start-up time. The pre-set timing can be considered as determined, for example, referring to the background art of the present application, it can be set during the Spring Festival holiday.

[0042] In step S20, the core items to be detected include but are not limited to whether the electrical parameters meet the operation requirements, whether the protection devices are normal, whether the temperature is in a normal state, and whether the physical connection between devices is correct and not loose; the non-core items to be evaluated include but are not limited to whether the environmental conditions affect the start-up of the device, whether the redundant device is normal, whether the cooling system is normal, and whether the ventilation device is normal.

[0043] Meanwhile, the fault prediction model of the present application adopts a long short-term memory network (LSTM) or an XGBoost algorithm, uses features in historical data as input, and evaluates the fault probability of the non-core items.

[0044] The fault rate threshold value can be manually set according to experience or dynamically adjusted by a dynamic adjustment model. The dynamic adjustment model is as follows:

[0045]

[0046] wherein θ represents the set fault rate threshold value; MTBF is the mean time between failures; R(t) represents the reliability evaluation function of a certain non-core item at the current time, R(t) = e -λt , λ represents the historical fault rate, which can be obtained by searching data; and α and β are weighting coefficients, which are manually set. For example, assuming that the mean time between failures MTBF of a certain non-core item of an SVG device is 10,000 hours, and the historical fault rate λ thereof is 0.0001 times / hour according to historical data analysis, the reliability evaluation result R(5000) at a certain time t = 5,000 hours is e -0.0001x5000 = 0.6065, which is brought into the above dynamic adjustment model formula, and the fault rate threshold value can be obtained.

[0047] In step S30, the output calculation parameter setting includes the following process: analyzing historical operation data by a machine learning algorithm, automatically matching the current power grid working condition, and dynamically generating the optimal output calculation parameter combination. The machine learning model of the output calculation parameter is as follows:

[0048] P * (X) = W T X + b;

[0049]

[0050] wherein, X represents an input feature vector, represents the relevant parameters of the power grid operating condition; W represents a weight vector, W T is the transpose of W; b represents a bias term, which is a constant value; P* represents the optimal output parameter setting value predicted by the model; n is the number of samples; L(W, b) is a loss function; X i is the feature vector of the i-th sample, Y i is the true value of the corresponding optimal output parameter.

[0051] In step S40, after the static var generator is started, the power grid load information is collected in real time, and the output power of the static var generator is dynamically adjusted in combination with the load change trend prediction model. By monitoring and predicting the change of the power grid load in real time, the output power of the SVG can be adjusted in advance to avoid system instability or low efficiency caused by sudden load change, and the stability of the power grid operation is ensured. In the embodiment, the load change trend prediction model adopts an ARIMA(p, d, q) model.

[0052] After the above scheme is adopted, the following effects can be achieved:

[0053] 1. Improve the response speed: by setting the SVG to a preset standby state in off-peak periods, the device can quickly switch from standby state to working state when the power grid load suddenly changes or needs to be adjusted urgently, greatly shortening the startup time;

[0054] 2. Optimize the detection process: distinguish between core items and non-core items, and use a fault prediction model based on historical operation data and machine learning algorithm to evaluate the failure rate of non-core items. In this way, while ensuring safety, unnecessary detection steps are reduced, further shortening the startup time;

[0055] 3. Dynamically adjust the failure rate threshold: use a dynamic adjustment model to calculate the reliability evaluation result at a certain moment according to the mean time between failures (MTBF) and historical failure rate, and dynamically adjust the failure rate threshold accordingly, improving the accuracy and flexibility of fault prediction;

[0056] 4. Intelligent output parameter setting: by analyzing historical operation data and using machine learning algorithm to automatically match the current power grid operating condition, the optimal output calculation parameter combination is dynamically generated, improving the output efficiency and adaptability of the SVG;

[0057] 5. Enhance system stability: after starting, the power grid load information is collected in real time, and the output power of the SVG is dynamically adjusted in combination with the load change trend prediction model, effectively avoiding system instability or low efficiency caused by sudden load change, and ensuring the stability and reliability of the power grid operation.

[0058] 6. Improve maintenance efficiency: By predicting the failure of non-core items, potential risks can be identified in advance, so that maintenance work can be planned, the possibility of sudden failure can be reduced, and maintenance cost and downtime can be reduced.

[0059] The above examples are only used to explain the concept of the present application, and are not limited to the protection of the present application. Any non-essential modification of the present application shall fall within the scope of the present application.

Claims

1. A quick start method for shortening the startup operation time of a static VAR generator, characterized in that: The following steps are involved: S10. In the off-peak operating period, the static VAR generator is configured to a preset standby state in advance, wherein the preset standby state includes setting the operating parameters of the static VAR generator; S20. Items related to the operation of the static VAR generator are divided into core items and non-core items. For core items, a comprehensive test is performed at startup. For non-core items, a fault prediction model built based on historical operating data and a machine learning algorithm is used to evaluate the failure rate of non-core items at the end or during the previous operating cycle. If the evaluation result of a non-core item is higher than the set failure rate threshold, the test is performed at startup. If the evaluation result of a non-core item is lower than the set failure rate threshold, the test is skipped at startup. S30. Output calculation parameter setting; S40. The static VAR generator startup is completed.

2. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: The core items that need to be tested include but are not limited to whether the electrical parameters meet the operating requirements, whether the protection devices are normal, whether the temperature is normal, and whether the physical connections between devices are correct and not loose; Non-core items that need to be evaluated include but are not limited to whether environmental conditions affect equipment startup, whether redundant equipment is functioning properly, whether the cooling system is functioning properly, and whether ventilation equipment is functioning properly.

3. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: The fault prediction model adopts the long short-term memory network LSTM or XGBoost algorithm, uses the features in the historical data as input, and evaluates the failure probability of non-core items.

4. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: The output calculation parameter setting includes the following process: Through machine learning algorithms, historical operating data is analyzed, the current grid conditions are automatically matched, and the optimal output calculation parameter combination is dynamically generated.

5. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: The machine learning model for the output calculation parameters is as follows: P * (X)=W T X+b; Where X represents the input feature vector, which represents the relevant parameters of the power grid operating conditions; W represents the weight vector, W T is the transpose of W; b represents the bias term, which is a constant value; P* represents the optimal output parameter setting value obtained by model prediction; n is the number of samples; L(W,b) is the loss function; X i is the feature vector of the i-th sample, Y i is the true value of the corresponding optimal output parameter.

6. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: In step S40, after the static VAR generator is started, grid load information is collected in real time, and the output power of the static VAR generator is dynamically adjusted in combination with a load change trend prediction model.

7. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 6, characterized in that: The load change trend prediction model adopts the ARIMA (p, d, q) model.

8. A fast startup method for shortening the startup operation time of a static VAR generator according to claim 1, characterized in that: The failure rate threshold is dynamically adjusted using a dynamic adjustment model, and the dynamic adjustment model is as follows: Where θ represents the set failure rate threshold; MTBF is the mean time between failures; R(t) represents the reliability evaluation function of a non-core item at the current moment; α and β are weighting coefficients, which are set manually.

Images