A substation temperature monitoring method based on improved Kalman filtering

Abstract

The application discloses a substation temperature monitoring method based on improved Kalman filtering, which comprises the following steps: measuring the temperature of a substation by using temperature sensors; establishing a state equation and an observation equation in the form of a temperature sensor matrix; detecting whether the measured data is lost by each sensor through a discriminant; preliminarily estimating the temperature measurement value by using a classical Kalman filter; calculating an optimal temperature estimation value in combination with the measured data of neighbor sensors; inputting the accurate temperature value into a PID control system, and controlling the operation of an air conditioner and an exhaust valve according to the monitoring requirement of the substation temperature, so that the substation temperature is controlled within a safe range. The application utilizes the improved Kalman filtering method and the PID control technology, effectively reduces the influence of the loss of the measured data on the real-time monitoring of the substation temperature, and controls the substation temperature within a safe range.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of temperature monitoring, and particularly relates to a substation temperature monitoring method based on improved distributed Kalman filtering. BACKGROUND

[0002] With the rapid development of high voltage, large unit and large capacity of electric power, whether the main electrical equipment in the substation can operate normally will directly relate to the operation of the power system. In the substation, the long-term normal live operation of the electrical equipment will generate a certain amount of heat, and the aging of the contact points and connecting points will cause the problem of increased contact resistance, which will also generate heat. If heat dissipation is not performed in time, the room temperature in the closed space will rise, and the running power electronic components, integrated boards and precision chips will be damaged, and even cause a fire and regional large-area power outage, causing a major power accident.

[0003] Kalman filter is a kind of efficient autoregressive filter, which can estimate the state of a dynamic system in the combined information of many uncertainties, and is a powerful and highly versatile tool. In essence, it is a data fusion algorithm that fuses data from different sensors with the same measurement purpose. The state space is used as a carrier to describe the estimation method, and the optimal estimate of the fused data in the statistical sense is recursively determined. The difference between the true value and the estimated value is called estimation error, and if the difference between the two is zero, it is called unbiased estimation. The optimal estimation of Kalman filter requires three conditions: unbiasedness, that is, the mean of the estimated value is close to the true value of the monitored object state; estimation variance minimization and real-time, that is, under the influence of random interference and noise, the optimal estimated value of the state of the monitored object is given in real time based on the linear minimum variance method. However, in actual application, because of the limitation of the battery energy of the sensor itself or the increase of the data volume, the measurement data often loses, so it becomes a practical topic to obtain unbiased estimation of temperature from measurement data and noise under intermittent observation. SUMMARY

[0004] In order to solve the above technical problems, the application provides a substation temperature monitoring method based on improved Kalman filtering, which can effectively cope with the influence of measurement data loss on accurate temperature estimation and perform real-time optimal monitoring of the temperature of the substation.

[0005] The technical scheme for solving the above technical problems is: a substation temperature monitoring method based on improved Kalman filtering, comprising the following steps:

[0006] S1, using a temperature sensor to measure the temperature of the substation in real time to obtain temperature data;

[0007] S2, according to the sensor system model, a temperature sensor state equation and an observation equation are established; the state equation and the observation equation in the form of a temperature sensor matrix are respectively:

[0008]

[0009]

[0010] wherein, and are the prior temperature prediction value of the sensor i at the k moment and the posterior temperature estimation value at the k-1 moment, A is a state transition matrix, υ i,k is system noise conforming to a Gaussian distribution; y i,k is an m-dimensional column vector composed of measurement data, p i,k is the probability that the measurement data of the sensor i at the k moment is not lost, H i,k is an m*n-dimensional measurement matrix, ω i,k is observation noise conforming to a Gaussian distribution;

[0011] S3, according to the observation data, the measurement data loss of the sensor is detected;

[0012] S4, according to the state equation and the observation equation and the loss, a classical Kalman filter is used to preliminarily estimate the temperature measurement value;

[0013] S5, combining the preliminary estimation and the system model, the optimal temperature estimation value is calculated after the measurement data of the neighbor sensor is fused;

[0014] S6, the optimal temperature value is input to a PID control system, and the start and stop of the air conditioner and the exhaust valve are set according to the monitoring requirements of the substation temperature, so that the substation temperature is controlled within a safe range.

[0015] Further, the statistical characteristic discriminant of whether the measurement data is lost in the step S3 is as follows:

[0016]

[0017] wherein, is the variance of the measurement vector y i,k when the m-dimensional measurement data of the sensor i at the k moment is not lost, is the variance of the measurement vector y i,k when the m-dimensional measurement data of the sensor i at the k moment is lost, when the discriminant ξ>0, the m-dimensional measurement data of the sensor i at the k moment is lost, and when the discriminant ξ<0, the m-dimensional measurement data of the sensor i at the k moment is normal.

[0018] Further, the step of step S4 of performing a preliminary estimation of the temperature measurement value by using a classical Kalman iteration is:

[0019] Temperature value prediction update:

[0020] Error covariance prediction update:

[0021] A is the predicted temperature error covariance of sensor i at time k, k B is the state transition matrix, k C is the measurement matrix,

[0022] Kalman gain calculation:

[0023] K is the calculated Kalman gain of sensor i at time k, A is the predicted temperature error covariance of sensor i at time k, i,k B is the state transition matrix, k C is the measurement matrix,

[0024] Temperature update estimation:

[0025] z is the posterior temperature estimation of sensor i at time k, i,k B is the state transition matrix, is calculated, wherein θ i,k is an indicator matrix, θ i,k = 0 when data loss occurs, and θ i,k = 1 when no data loss occurs; when data loss occurs, the corresponding data of the original measurement is discarded and replaced with the local prior estimation at time k for the Kalman iteration operation;

[0026] Error covariance update:

[0027] and are the posterior temperature error covariance and the predicted prior temperature error covariance of sensor i at time k, K is the calculated Kalman gain of sensor i at time k.

[0028] Further, the step of step S5 of calculating the optimal temperature estimation value in combination with the measurement data of the neighbor sensors is:

[0029]

[0030] wherein, is the optimal estimation value of the local temperature of the sensor i at the k time after fusing the data, w ij is a weight matrix, N i is the neighbor set of the sensor i.

[0031] The technical effect of the present application is that the improved distributed Kalman filtering method can obtain a sub-error substation temperature estimation value from the measurement data, system noise and measurement noise compared with the traditional Kalman filtering algorithm through detection, estimation and fusion of the measurement data, and the accurate temperature estimation value is input into the PID control system, and the start and stop of the air conditioner and the exhaust valve are set in a deviation-eliminating manner according to the monitoring requirements of the substation temperature by means of the relay output module, so as to ensure that the substation temperature fluctuates within a safe range. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a flowchart of the present application;

[0033] Figure 2 is a flowchart of the improved Kalman algorithm in the present application;

[0034] Figure 3 is a simulation diagram of temperature monitoring in the present application using the improved Kalman algorithm;

[0035] Figure 4 is an error comparison diagram using the improved Kalman algorithm in the present application;

[0036] Figure 5 is a closed-loop logic control diagram in the present application. DETAILED DESCRIPTION

[0037] The technical solutions of the present application will be described in detail below in combination with the drawings and specific embodiments.

[0038] As shown in Figure 1 , the present application provides a substation temperature monitoring method based on an improved Kalman filter, which specifically comprises the following steps:

[0039] S1, using a temperature sensor to measure the temperature of the substation in real time to obtain temperature data;

[0040] A plurality of temperature sensors are arranged to preliminarily measure the temperature data in the station, and the obtained measurement data includes temperature signals and noise, and the measurement values are in a discrete distribution state. As shown in Figure 2 , the temperature data measured from the temperature sensor is transmitted to the improved Kalman algorithm through a digital information input module.

[0041] S2, establishing a temperature sensor state equation and an observation equation according to a sensor system model;

[0042] Based on the mathematical characteristics of temperature sensor data, the state equation in matrix form is established as follows:

[0043]

[0044] in, and Let A be the prior temperature prediction of sensor i at time k and the posterior temperature estimate at time k-1, respectively, and let A be the state transition matrix. i,k ~N(0,R k To address system noise conforming to a Gaussian distribution, the improved Kalman oscillator mentioned in this invention treats unfused temperature values ​​as local temperatures, denoted by loc in the superscript of state variables. Conversely, fused temperature values ​​are treated as local temperatures, denoted by reg in the superscript of state variables. Data communication is limited to neighboring temperature sensors, thus minimizing computational and communication burdens on the system.

[0045] The observation equation in matrix form is established as follows:

[0046]

[0047] Among them, y i,k The measurement data is represented by an m-dimensional column vector. Let p be the prior temperature prediction of sensor i at time k. i,k The probability that sensor i has not lost its measurement data at time k can be set by prior knowledge in practical applications. i,k Let ω be an m*n dimensional measurement matrix. i,k ~N(0,Q) k ) represents observation noise that conforms to a Gaussian distribution.

[0048] Figure 2 This is a flowchart illustrating the improved Kalman algorithm in this invention. In this case, S3, S4, and S5 represent one Kalman cycle, from which the optimal temperature estimate at time k can be obtained. In practical applications, before the first Kalman filtering cycle begins, it is necessary to adjust relevant parameters, such as the noise error covariance R of the system model, according to the system model of the temperature sensor system. k Measure the noise covariance Q k Initial temperature value t i,0 Initial temperature covariance The data is then input into a Kalman filter to monitor the temperature inside the substation in real time.

[0049] S3, based on the observation data, detects the loss of sensor measurement data;

[0050] Based on the statistical characteristics of the measurement data before and after loss, the discriminant for detecting data loss is as follows:

[0051]

[0052] If the discriminant ξ > 0, the mth dimension measurement data of sensor i at time k is lost.

[0053] If the discriminant ξ < 0, the mth dimension measurement data of sensor i at time k is normal.

[0054] wherein, is the variance of the measurement vector y i,k if the mth dimension measurement data of sensor i at time k is not lost. is the variance of the measurement vector y i,k if the mth dimension measurement data of sensor i at time k is lost.

[0055] S4, according to the system model, the state equation, the observation equation and the loss condition, the temperature measurement value is preliminarily estimated by the classical Kalman filtering;

[0056] Kalman filtering is an algorithm for optimal estimation of a system by using linear state equation and observation data of system input and output. Since the observation data contains noise and interference, the optimal estimation can also be regarded as a filtering process. Kalman filtering is divided into two steps of prediction update and measurement update.

[0057] Each sensor preliminarily estimates the temperature measurement value by classical Kalman iteration, which has the following steps:

[0058] Temperature value prediction update:

[0059] Error covariance prediction update:

[0060] is the predicted temperature error covariance of sensor i at time k, A k is the state transition matrix, Q k is the covariance matrix of measurement noise.

[0061] Kalman gain calculation:

[0062] is the calculated Kalman gain of sensor i at time k, is the predicted temperature error covariance of sensor i at time k, H i,k is the measurement matrix, Q k is the covariance matrix of measurement noise.

[0063] Temperature measurement update:

[0064] is the posterior temperature estimate of sensor i at time k, z i,k is the reconstructed measurement matrix, which is calculated by is calculated, where θ i,k is an indicator matrix, θ i,k = 0 when data loss occurs, and θ i,k = 1 when no data loss occurs. If data loss occurs, the corresponding entry of the original measurement will be discarded and replaced by the local prior estimate at time k for the Kalman iteration.

[0065] Error covariance measurement update:

[0066] and are the posterior temperature error covariance and the predicted prior temperature error covariance of sensor i at time k, respectively, is the Kalman gain calculated for sensor i at time k. After this step, each sensor can obtain the optimal temperature estimate available to the local sensor by eliminating the effect of errors through known measurement data.

[0067] S5, combining the preliminary estimate and the system model, calculates the optimal temperature estimate value after fusing the measurement data of the neighbor sensors;

[0068] Combining the measurement data of the neighbor sensors, the optimal temperature estimate value is calculated:

[0069]

[0070] wherein, is the local optimal temperature estimate value obtained by sensor i at time k after fusing data, w ij is a weight matrix, and a consensus weight matrix such as the maximum degree weight method and the Metropolis weight method is usually used. N i is the neighbor set of sensor i. Each sensor performs fusion using only information obtained from locally connected neighbors, which can provide enhanced redundant information for each sensor node, thereby improving the robustness against sensor failure and improving the estimation accuracy.

[0071] Figure 3 is a simulation schematic diagram of temperature monitoring in the substation temperature control method using the improved Kalman algorithm. In the simulation experiment, the ideal temperature of the substation is 25 degrees, and the actual temperature fluctuates up and down around 25 degrees. Six temperature sensors are simulated in a device operating room for real-time monitoring, and the measurement data loss probability of each temperature sensor is set to 0.1. The simulation results are shown in Figure 3 , and the six sensors can accurately estimate the temperature from the data containing measurement noise.

[0072] Figure 4 is the error comparison chart of the improved Kalman algorithm in the substation temperature control method.

[0073] The error measurement standard adopts the root mean square error, and the measurement error, the error result based on the improved centralized Kalman algorithm estimation error and the improved distributed Kalman algorithm is as shown in Figure 4 The error of the improved distributed Kalman algorithm used in the application is the smallest, the accuracy is the highest, and the influence of data loss on the estimation accuracy can be effectively reduced.

[0074] S6, the accurate temperature value is input to the PID control system, the start and stop of the air conditioner and the exhaust valve are set according to the monitoring requirement of the substation temperature, so that the substation temperature is controlled in the safe range.

[0075] An accurate temperature estimation value is obtained through data fusion, as shown in Figure 5 The accurate temperature value is input to the PID control system, the start and stop of the air conditioner and the exhaust valve are controlled in the form of eliminating deviation according to the monitoring requirement of the substation temperature by means of the relay output module, when the temperature is in the safe range, the PID stops outputting, so that the indoor temperature is basically kept constant.

[0076] It is worth noting that the average heat generation of different equipment in the substation also has differences (for example, the heat generation of GIS equipment is 85kW / h, and the average heat generation of the integrated computer system is only 3.3kW / h), so the local temperature in the station will also be weak. In most cases, the temperature control device is also set near the equipment with large heat generation, so the temperature difference in the station can be approximately equal, but in some special cases, there will be a larger difference, so the measurement accuracy of the temperature sensor distributed in each direction after Kalman filtering iteration is not as high as the estimation value after data fusion, but it reflects the local temperature and should also be paid attention to.

[0077] The specific embodiments of the application are described above. It should be understood that the application is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essential content of the application. In the case of no conflict, the embodiments of the application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A substation temperature monitoring method based on improved Kalman filtering, characterized in that: S1, using a temperature sensor to measure the temperature of the substation in real time to obtain temperature data; S2, according to the sensor system model, establishing a temperature sensor state equation and an observation equation; the state equation and the observation equation of the temperature sensor in matrix form are respectively: S3, according to the observation data, detecting the measurement data loss of the sensor; S4, according to the state equation and the observation equation and the loss, using a classical Kalman filter to preliminarily estimate the temperature measurement value; S5, combining the preliminary estimate and the system model, calculating the optimal temperature estimate value after fusing the measurement data of the neighbor sensor; S6, inputting the optimal temperature value into a PID control system, setting the start and stop of the air conditioner and the exhaust valve according to the monitoring requirements of the substation temperature, and controlling the substation temperature within a safe range. The statistical characteristic discriminant of whether the measurement data is lost in the step S3 is as follows: The step of preliminarily estimating the temperature measurement value in the step S4 using a classical Kalman iteration is as follows: wherein, and are the prior temperature prediction value of sensor i at time k and the posterior temperature estimate value at time k-1, respectively, A is the state transition matrix, υ i,k is the system noise following a Gaussian distribution; y i,k is an m-dimensional column vector composed of the measurement data, p i,k is the probability of the measurement data of sensor i at time k not being lost, H i,k is an m*n-dimensional measurement matrix, ω i,k is the observation noise following a Gaussian distribution; The optimal temperature estimate value is calculated in the step S5 by combining the measurement data of the neighbor sensor: ​ ​ ​ 2. The substation temperature monitoring method based on modified Kalman filtering according to claim 1, characterized in that, ​ wherein, is the variance of the measurement vector y i,k is the variance of the measurement vector y i,k When the discriminant ξ > 0, the mthdimensional measurement data of sensor i at time k is missing, and when the discriminant ξ < 0, the mthdimensional measurement data of sensor i at time k is normal.​ 3. The modified Kalman filter based substation temperature monitoring method of claim 1, wherein, ​ Temperature value prediction update: Error covariance prediction update: A is the temperature error covariance predicted for sensor i at time k k Q is the state transition matrix k R is the covariance matrix of the observation noise Kalman gain computation: Kalman gain computed for sensor i at time k, temperature error covariance predicted for sensor i at time k, H i,k measurement matrix, Q k covariance matrix of the measurement noise; Temperature update estimate: is the posterior temperature estimate of sensor i at time k, z i,k is the reconstructed measurement matrix, given by the formula is calculated, where θ i,k is an indicator matrix, θ i,k = 0 when data loss occurs, and θ i,k = 1 when no data loss occurs; when data loss occurs, the corresponding data of the original measurement will be discarded and replaced by the local prior estimate at time k for the Kalman iteration operation; Error covariance update: and are the sensor i posterior temperature error covariance at time k and the predicted prior temperature error covariance, respectively, is the Kalman gain computed by sensor i at time k.

4. The modified Kalman filter based substation temperature monitoring method of claim 1, wherein, ​ where, is the local temperature optimal estimation of sensor i after fusing data at time k, w ij is a weight matrix, N i is the neighbor set of sensor i.

Images