Multi-sensor data fusion method for intelligent transfer device of power equipment

Abstract

The invention discloses a multi-sensor data fusion method for an intelligent transfer device of power equipment, and the method comprises the steps: carrying out the dynamic weight fusion of multi-source data through building a measurement model of a plurality of visual sensors, and generating a high-precision measurement sequence; in combination with an improved adaptive Kalman filtering algorithm, process noise and a measurement noise covariance matrix are estimated in real time, and the stability and the anti-interference capability of state estimation are optimized. And based on the fused high-precision state information, a real-time path planning instruction is generated, and the transfer device is controlled to complete obstacle avoidance and millimeter-level docking of the cabinet body. The problems that traditional manual transfer efficiency is low, the positioning error is large, and a single sensor is prone to electromagnetic interference are solved, the navigation precision and operation safety of heavy equipment in the complex electromagnetic environment are remarkably improved, and the method is suitable for intelligent transfer scenes of equipment such as transformer substation circuit breaker handcarts and the like.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing and automatic control technology, and in particular to the technical field of a multi-sensor data fusion method for an intelligent power equipment transfer device. Background Technology

[0002] Currently, substation operation and maintenance in my country faces significant technical bottlenecks: during the overhaul of circuit breakers at voltage levels of 35kV and above, the transfer of equipment using handcarts still relies on manual operation. Traditional operating methods face three major challenges: 1) Handcarts typically weigh over 800kg, posing a risk of falling during manual handling; 2) Cabinet docking precision requirements reach millimeter level, making it difficult for manual operation to guarantee accurate positioning; 3) The complex electromagnetic environment of substations makes single sensors susceptible to interference, leading to navigation failure. This situation not only affects the efficiency of maintenance work but also poses significant safety hazards, hindering the orderly progress of maintenance work. To improve the efficiency of circuit breaker overhaul preparation while ensuring worker safety during handcart transfer, it is necessary to develop and design a dedicated transfer device. This new transfer vehicle should possess automated and intelligent functions to assist workers in equipment handling, docking, lifting, and placement, thereby achieving safe and stable transfer of the handcart. An ideal transfer device should be adaptable to different models of circuit breaker handcarts, capable of precise docking with cabinets, stable support for heavy-duty objects, and able to be operated by a single person to complete the entire transfer process.

[0003] This innovative design not only significantly improves transfer efficiency and reduces workers' workload, but also greatly reduces the risk of accidental injuries and equipment damage caused by human factors during operations. The introduction of this new transfer vehicle will help substations achieve a higher level of standardization in daily operation and maintenance, promoting a comprehensive improvement in the safety, stability, and efficiency of the power system, thus laying a solid foundation for smart grid construction.

[0004] However, for heavy power equipment transfer tasks, precise operation is crucial. It is necessary to apply multi-sensor information fusion technology to the navigation and path planning of autonomous mobile platforms. By integrating data from multiple different measurement sources, it is possible to achieve better results and greater depth than a single information source. Existing solutions have significant drawbacks: traditional Kalman filtering algorithms are prone to divergence when the noise statistics are unknown, leading to the accumulation of navigation and positioning errors. Therefore, this invention integrates multi-sensor information perception and adaptive filtering technology, proposing a noise-robust data fusion method to provide high-precision state estimation capabilities for intelligent transfer devices. Summary of the Invention

[0005] The purpose of this invention is to solve the problems in the prior art and propose a multi-sensor data fusion method for intelligent power equipment transfer devices. This method enables the transfer devices to achieve millimeter-level positioning accuracy in complex electromagnetic environments, suppresses navigation error accumulation through adaptive noise covariance matrix estimation, enhances anti-interference capabilities by fusing multi-source sensor data, and reduces the risk of manual operation while improving the safety and efficiency of heavy equipment transfer.

[0006] To achieve the above objectives, this invention proposes a multi-sensor data fusion method for an intelligent power equipment transfer device, which specifically includes the following steps: S1. Establish sensor measurement models for multiple cameras; S2. Measurement sequence fusion: By fusing measurement data from multiple cameras, weighted or combined measurement values ​​are obtained, thereby improving the accuracy of the measurement data. S3, Process noise covariance matrix estimate; S4, Measurement noise covariance matrix estimate; S5. Based on the above steps of measurement sequence fusion and noise covariance matrix estimation, this invention provides a multi-sensor data fusion method based on improved adaptive Kalman filtering to achieve high-precision estimation of state vectors. S6. Based on the high-precision state estimation results, the real-time motion command of the transfer device is generated through the path planning algorithm, and the chassis drive module is controlled to complete obstacle avoidance and millimeter-level docking with the cabinet, ensuring that the positioning error is less than 1mm.

[0007] In step S2, the different components of the same target measured by each sensor are processed, and the measurement data of each sensor are gradually fused to obtain a fused measurement sequence. During the fusion process, a weighted average value is calculated based on the measurement variance of each sensor to improve the accuracy of the fused data.

[0008] In step S3, a linear Kalman filter model is considered, based on the time from the initial time to the time... Based on the measurement sequence, a new measurement model is calculated, a noise sample sequence is constructed, and the process noise covariance matrix is ​​estimated in real time using the law of large numbers and recursive estimation methods.

[0009] In step S4, the measurement noise covariance matrix is ​​adaptively estimated by utilizing the relationship between the innovative sequence and the measurement noise covariance matrix.

[0010] In step S5, the measurement sequence fusion adaptive Kalman filter algorithm is implemented by: initializing the state vector and covariance matrix; updating the Kalman gain matrix at each time step based on the fused measurement sequence and innovation sequence; and estimating the state vector using the updated Kalman gain matrix.

[0011] This invention also proposes an intelligent transfer device for power equipment, which uses the aforementioned multi-sensor data fusion method for real-time positioning and path planning. This invention further proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned multi-sensor data fusion method.

[0012] The specific process of measurement sequence fusion in step S2 is as follows: Will Measurement data from each sensor are fused component by component, where the first... During the second fusion, according to the first The measurement variance of each sensor is compared with the variance of the previous fusion result, and a weighted average is calculated. After the first fusion, the fused measurement sequence is obtained, where the first... The weighting coefficient for the sub-fusion is , For the first Measurement variance of each sensor, This represents the variance of the previous fusion result.

[0013] The recursive estimation process noise covariance matrix described in step S3 include: A sample sequence containing process noise and measurement noise is constructed. Based on a linear Kalman filter model, the process noise covariance matrix is ​​calculated in real time using the law of large numbers and recursive estimation methods. The estimated value.

[0014] The adaptive estimation of the measurement noise covariance matrix in step S4 The specific method is as follows: based on the statistical characteristics of the innovation sequence, through the formula... Calculate the measurement noise covariance matrix The real-time estimated value.

[0015] The improved adaptive Kalman filter algorithm described in step S5 includes: initializing the state vector and error covariance matrix; updating the Kalman gain matrix according to the fused measurement sequence; and iteratively performing state prediction, gain calculation, state update and covariance update to finally output a high-precision state estimate.

[0016] The method is applied to the substation circuit breaker trolley transfer scenario, achieving millimeter-level positioning accuracy through multi-sensor data fusion and adapting to noise interference in complex electromagnetic environments.

[0017] The multiple cameras include at least three visual sensors, each equipped with an electromagnetic interference shielding layer, operating in the 1-5GHz frequency band, and sampling at a frequency of not less than 30Hz, used to collect environmental information of the transport device during navigation and obstacle avoidance.

[0018] The method, by fusing data from multiple sensor sources, achieves a root mean square error smaller than that of the traditional linear Kalman filter algorithm, and improves the stability of navigation signal tracking.

[0019] The present invention also proposes an intelligent transfer device for power equipment, which uses the above-mentioned method for real-time positioning and path planning.

[0020] The present invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0021] The beneficial effects of this invention are: Improved filtering accuracy: By fusing measurement data from multiple sensors, the accuracy of the measurement data is effectively improved, thereby improving the estimation accuracy of the filter.

[0022] Enhanced stability: By accurately estimating the noise covariance matrix, the filter divergence problem caused by unknown noise statistical characteristics is avoided, thus enhancing the stability of the filter.

[0023] Highly adaptable: Suitable for various noise environments and sensor configurations, exhibiting strong adaptability.

[0024] Optimized navigation signal tracking: In navigation signal tracking applications, this algorithm can track signals quickly and accurately, significantly improving the performance of the navigation system.

[0025] Experimental results show that when the electromagnetic interference intensity is ≥50dBμV / m, the root mean square error of positioning using this method is less than 1.5mm, the path tracking delay is less than 100ms, and the accuracy is improved by about 40% compared with the traditional linear Kalman filter (LKF).

[0026] The features and advantages of the present invention will be described in detail through embodiments and in conjunction with the accompanying drawings. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the algorithm of a multi-sensor data fusion method for an intelligent power equipment transfer device according to the present invention. Figure 2 This is a performance comparison chart of the algorithm with other algorithms in an embodiment of the multi-sensor data fusion method for an intelligent power equipment transfer device of the present invention; Figure 3This is a root mean square error diagram between the algorithm of the multi-sensor data fusion method of the intelligent transfer device for power equipment of the present invention and the estimated value and the real value obtained by other algorithms; Figure 4 This is a diagram showing the result of multi-sensor data fusion, which is the algorithm of the multi-sensor data fusion method for an intelligent power equipment transfer device according to the present invention. Detailed Implementation

[0028] See Figure 1 , Figure 2 , Figure 3 and Figure 4 The method of the present invention includes the following steps: S1. Establish sensor measurement models for multiple cameras; S2. Perform measurement sequence fusion on the measurement data from multiple cameras, obtain the weighted average value by calculating the measurement variance of each sensor, and generate the fused measurement sequence; S3. Based on the measurement sequence from the initial time to the current time, recursively estimate the process noise covariance matrix Q; S4. Using the relationship between the innovative sequence and the measurement noise covariance matrix, adaptively estimate the measurement noise covariance matrix R; S5. Based on the improved adaptive Kalman filter algorithm, combined with the fused measurement sequence and process noise covariance matrix... and measurement noise covariance matrix High-precision estimation of the state vector of the transfer device is achieved, enabling multi-sensor data fusion; S6. Based on the high-precision state estimation results of step S5, generate real-time path planning instructions for the transfer device, and control the transfer device to complete obstacle avoidance and millimeter-level docking with the cabinet.

[0029] This invention also proposes an intelligent transfer device for power equipment, which uses the aforementioned multi-sensor data fusion method for real-time positioning and path planning. This invention further proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned multi-sensor data fusion method.

[0030] The working process of this invention: The multi-sensor data fusion method of the intelligent transfer device for power equipment of the present invention is described in conjunction with the accompanying drawings during operation.

[0031] This embodiment considers the real-time positioning and obstacle avoidance operations of the transport vehicle in complex environments. It provides a multi-sensor data fusion method based on measurement sequence fusion and adaptive Kalman filtering. This method fuses data from multiple camera sensors on the transport vehicle to improve its positioning accuracy and environmental perception. The multiple cameras include at least three visual sensors, each equipped with an anti-electromagnetic interference shielding layer. The visual sensors use a metal shielding shell, operate in the 1-5GHz frequency band, and have a sampling frequency of no less than 30Hz to suppress high-frequency electromagnetic interference within the substation. This data is used to collect environmental information during the navigation and obstacle avoidance process of the transport device. Figure 1 As shown, it includes the following steps: S1. Establish sensor measurement models for multiple cameras.

[0032] in, For the first The measurement vector at time (observations from one or more sensors); The observation matrix (the matrix that maps the state vector to the measurement space); For the first The state vector at any given moment (such as the position and velocity of the transfer device); For the first The measurement noise vector at time (following a zero-mean normal distribution, with a covariance matrix of...) ).

[0033] First, consider data fusion from a single radar data source. Let's assume a single data source... have If we want to use these observations to obtain a true value... If the estimated value is [value], then the average value is taken. Assume [the value]. ,so For the first The single-sensor measurement value at time t. According to the above formula, it can be observed that as... The value gradually increases. It gets smaller and smaller. This means that when there is a large amount of data, the impact of the current measurement on the estimate is less significant. Conversely, when... When the data is relatively small (few data points), it relies more on the current measurements. Let Then the above formula can be written as: Weighting coefficients (depending on the amount of data) (Increases and decreases reflect dependence on historical data). If two sensors are used simultaneously to acquire measured values, the standard deviations of the measured values ​​from these two sensors are respectively... , Furthermore, these measured values ​​conform to a normal distribution. Two sensors simultaneously measure the same data. The values ​​are respectively and .according to and Real values ​​can be obtained The estimated value : if only variance The smaller, the better The closer the value is to the real value, the more: The two ends of the above equation are opposite Differentiating, we get: The above is a data fusion of data measured by two sensors.

[0034] S2, Measurement sequence fusion. Now assume there is... There are 1 sensor, of which: It is the first One sensor in The time measured at the first moment Each component can be fused using the method in S1. conduct The data fusion process involves first fusing the data from the first two sensors, then fusing the fused data with the data from the third sensor, and so on, until the data from the third sensor is fused with the data from the fourth sensor. Data from multiple sensors is fused and processed in total. (Secondary fusion) data fusion to obtain A matrix of order 1. Therefore: in, The total number of sensors, For the first The sensor at the first The first moment Each measurement component For the first The weighting coefficients of the next fusion (based on the variance of the previous fusion) and current sensor variance ), For the first The component difference between the second fusion and the previous fusion (used to calculate the weighted sum).

[0035] Indicates the first Measurement variance of each sensor, This represents the variance of the previous fusion result. Indicates the first The sensor measurement variance after data fusion, the first The weighting coefficient for the sub-fusion is By dynamically adjusting the weights, sensor data with smaller variance is selected first, thereby improving the fusion accuracy.

[0036] Indicates in At that moment, the The sensor after the second data fusion and the first The sensor after the second data fusion The difference between the components. When hour, , Indicates the first The first data fusion Each component.

[0037] S3, Process noise covariance matrix estimate.

[0038] Consider the following model: in For the sample time, The state transition matrix (describing the transition from state to state) arrive (changes) The input matrix (maps process noise to the state space). For state vectors, This is the measurement vector. It is the observation matrix. It is the input matrix. and Let process noise and measurement noise, respectively, be zero-mean process noise and measurement noise, both following a zero-mean normal distribution. Their covariance matrices are respectively... and In this model, based on the time from the initial time to time... A new measurement model is calculated from the measurement sequence: In the formula Indicates the initial state value of the system; Represents the sensor measurement sequence. Represents the observable matrix of the system. This represents the extended input matrix. The right-hand side of the equation can be viewed as two parts; the first part is the initial state value of the system. The second part is the noise sequence. and .but The value is generally not precise enough, so To some extent, this can be seen as an unknown disturbance.

[0039] Consider matrix , This is an auxiliary matrix used to eliminate the initial state. The impact on noise estimation follows ,in Let be the observable matrix of the system. Let be the row number of the auxiliary matrix, then Thus, the equation only contains noise sequences, because and It is white noise with zero mean and no correlation between the noise particles, therefore , They are also unrelated.

[0040] for Its sample sequence is .

[0041] in, The sample sequence contains only measurements of the noise term. Introducing symbols , , .make in, This serves as an intermediate variable (the basis for estimating the noise covariance). so in addition, , For matrix The sum of row vectors (used to solve) (generalized inverse), when The rank is Then, a generalized inverse exists. satisfy So Based on this, estimation can be performed in real time based on the system's measured values. ,by As The estimated value, As The real-time calculated value can ultimately yield: S4, Measurement noise covariance matrix estimate.

[0042] Considering the unknown precise statistical measurement noise covariance matrix With innovation sequence There is a certain relationship: in, The prediction error covariance matrix (an intermediate variable in Kalman filtering); For linear time-invariant systems, the covariance matrix of the innovative sequence can be obtained. The following estimates: in, For innovative sequences (the difference between measured and predicted values); Will The recursive estimate (based on historical innovation sequences) Substituting into the above equation, we can obtain the covariance matrix. The estimated value: S5. Adaptive Kalman Filtering Algorithm. The application of LKF (Traditional Linear Kalman Filter) is severely limited by the statistical information of additive noise. Although MSAKF (Measurement Sequence Adaptive Kalman Filter) solves this problem, the accuracy of the algorithm needs optimization. Based on this, a multi-sensor measurement sequence fusion adaptive Kalman filter is proposed to address this issue. The algorithm flow is as follows: Step 5.1 Initialize parameters, including system model parameters: state transition matrix Observation matrix Input matrix Initial state error covariance matrix Initial state value Initial matrix .

[0043] Step 5.2 Calculate the fusion measurement sequence according to step S2 .

[0044] Step 5.3 Enters the loop processing, calculating based on steps S3 and S4. , , .

[0045] Step 5.4 State prediction: Based on the system model parameters and the state estimate from the previous time step, the state at the current time step is predicted. The influence of process noise is considered when calculating the prediction error covariance. in, For the first The predicted state vector at time step; Step 5.5 Kalman gain calculation, used to weigh the predicted state and the actual measured value in the measurement update step. The Kalman gain matrix (weighing the prediction and measurement weights) is as follows: Step 5.6 State Update: Use Kalman gain and measurements to update the state estimate and reduce prediction error. in, The fused measurement sequence (from step S2); Step 5.7 Updates the error covariance to reflect the uncertainty of the updated state estimate: in, It is the identity matrix; Step 5.8 Wait for the sensor data at the next moment and repeat steps 5.3 to 5.7.

[0046] Ultimately, algorithm performance is compared to... Figure 2 As shown, it can be noted that the algorithm proposed in this patent outperforms LKF in signal tracking even in the presence of unknown noise. The root mean square error between the estimated value and the real value obtained by the algorithm is as follows: Figure 3 As shown, the root mean square error of this algorithm is slightly smaller than that of LKF. Figure 4The algorithm was demonstrated to achieve data fusion from three radar sensors, and the fusion result can track the true values ​​relatively well. In conclusion, simulation experiments prove the effectiveness of the method proposed in this invention.

[0047] The transfer device includes an omnidirectional wheel chassis, a multi-sensor array (including 3 visual sensors, 1 16-line lidar and 9-axis IMU) and an embedded control unit (equipped with an ARM Cortex-A72 processor). Through real-time data fusion and path planning, it achieves autonomous navigation under a load of 800kg.

[0048] The above embodiments are illustrative of the present invention and are not intended to limit the present invention. Any simple modifications to the present invention are within the scope of protection of the present invention.

Claims

1. A multi-sensor data fusion method for an intelligent power equipment transfer device, characterized in that: Includes the following steps: S1. Establish sensor measurement models for multiple cameras; S2. Perform measurement sequence fusion on the measurement data from multiple cameras, obtain the weighted average value by calculating the measurement variance of each sensor, and generate the fused measurement sequence; S3. Based on the measurement sequence from the initial time to the current time, recursively estimate the process noise covariance matrix Q; S4. Using the relationship between the innovative sequence and the measurement noise covariance matrix, adaptively estimate the measurement noise covariance matrix R; S5. Based on the improved adaptive Kalman filter algorithm, combined with the fused measurement sequence and process noise covariance matrix... and measurement noise covariance matrix High-precision estimation of the state vector of the transfer device is achieved, enabling multi-sensor data fusion; S6. Based on the high-precision state estimation results of step S5, generate real-time path planning instructions for the transfer device, and control the transfer device to complete obstacle avoidance and millimeter-level docking with the cabinet.

2. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The specific process of measurement sequence fusion in step S2 is as follows: Will Measurement data from each sensor are fused component by component, where the first... During the second fusion, according to the first The measurement variance of each sensor is compared with the variance of the previous fusion result, and a weighted average is calculated. After the first fusion, the fused measurement sequence is obtained, where the first... The weighting coefficient for the sub-fusion is , For the first Measurement variance of each sensor, This represents the variance of the previous fusion result.

3. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The recursive estimation process noise covariance matrix described in step S3 include: A sample sequence containing process noise and measurement noise is constructed. Based on a linear Kalman filter model, the process noise covariance matrix is ​​calculated in real time using the law of large numbers and recursive estimation methods. The estimated value.

4. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The adaptive estimation of the measurement noise covariance matrix in step S4 The specific method is as follows: based on the statistical characteristics of the innovation sequence, through the formula... Calculate the measurement noise covariance matrix The real-time estimated value.

5. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The improved adaptive Kalman filter algorithm described in step S5 includes: initializing the state vector and error covariance matrix; updating the Kalman gain matrix according to the fused measurement sequence; and iteratively performing state prediction, gain calculation, state update and covariance update to finally output a high-precision state estimate.

6. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The method is applied to the substation circuit breaker trolley transfer scenario, achieving millimeter-level positioning accuracy through multi-sensor data fusion and adapting to noise interference in complex electromagnetic environments.

7. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The multiple cameras include at least three visual sensors, each equipped with an electromagnetic interference shielding layer, operating in the 1-5GHz frequency band, and sampling at a frequency of not less than 30Hz, used to collect environmental information of the transport device during navigation and obstacle avoidance.

8. The multi-sensor data fusion method for an intelligent power equipment transfer device according to claim 1, characterized in that: The method, by fusing data from multiple sensor sources, achieves a root mean square error smaller than that of the traditional linear Kalman filter algorithm, and improves the stability of navigation signal tracking.

9. An intelligent transfer device for power equipment, characterized in that, Real-time positioning and path planning are performed using the multi-sensor data fusion method described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-sensor data fusion method according to any one of claims 1 to 8.

Images