Test method, test system and storage medium

Abstract

The invention relates to the technical field of power consumption testing, in particular to a testing method, a testing system and a storage medium. The test method comprises the following steps: obtaining an operation log of the intelligent lock in a preset time period; identifying a working state sequence of the intelligent lock based on the operation log; collecting real-time power consumption data corresponding to each working state, and adaptively and dynamically adjusting the sampling frequency; establishing a correlation model of the working state and the power consumption of the intelligent lock by utilizing a machine learning algorithm; calculating the total power consumption of the intelligent lock in a preset time period based on the correlation model; acquiring environmental factor data, and performing correction calculation to obtain total correction power consumption. The power consumption change of the intelligent lock in different working states can be more accurately captured by adaptively and dynamically adjusting the sampling frequency.

Description

Technical Field

[0001] The present invention relates to the technical field of power consumption testing, and in particular to a testing method, a testing system and a storage medium. Background Art

[0002] Smart locks have been widely used in the field of smart home security due to their convenience of being free from physical keys. However, as a battery-powered device, the power consumption of smart locks directly affects their service life and user experience.

[0003] The traditional power consumption test method of smart locks mainly relies on sampling at fixed time intervals and simple data accumulation. This method is difficult to adapt to the power consumption variation characteristics of smart locks in different working states. Since the working modes of smart locks are complex, including multiple states such as standby, identification, unlocking, and locking, the power consumption characteristics of each state are different. Fixed sampling frequency can easily lead to data distortion or omission of key information. In addition, environmental factors such as temperature and humidity also have a certain impact on the power consumption of smart locks, but existing methods often ignore these factors, resulting in a lack of comparability of test results. Summary of the invention

[0004] The present invention provides a testing method, a testing system and a storage medium to solve the technical problem of insufficient accuracy in power consumption testing of existing smart locks.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: In one aspect, a testing method is provided, the steps of the testing method comprising: Get the operation log of the smart lock within the preset time period; Identify the working state sequence of the smart lock based on the operation log; Collect real-time power consumption data corresponding to each working state and dynamically adjust the sampling frequency adaptively; Use machine learning algorithms to establish a correlation model between the working status and power consumption of smart locks; Calculate the total power consumption of the smart lock within a preset time period based on the association model; The environmental factor data is obtained and the total corrected power consumption is obtained by correction calculation.

[0006] On the other hand, a test system is provided for testing the power consumption of a smart lock, the test system comprising: The log acquisition module is used to obtain the operation log of the smart lock within a preset time period; A state recognition module is used to identify the working state sequence of the smart lock based on the operation log; The power consumption collection module is used to collect the real-time power consumption data corresponding to each working state according to the dynamically adjusted sampling frequency; A correlation model building module is used to build a correlation model between the working state and power consumption of the smart lock using a machine learning algorithm; A power consumption calculation module, used to calculate the total power consumption of the smart lock within a preset time period based on the association model; The environmental correction module is used to obtain environmental factor data and correct the calculated total power consumption.

[0007] On the other hand, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor of a computer, the above-mentioned testing method is implemented.

[0008] The beneficial effects of the present invention are: The present invention can more accurately capture the changes in power consumption of smart locks under different working states by adaptively and dynamically adjusting the sampling frequency. A correlation model between working state and power consumption is established using a machine learning algorithm to achieve fast and efficient power consumption prediction. An environmental factor correction mechanism is introduced to consider the effects of factors such as temperature, humidity, light intensity, and atmospheric pressure on power consumption, making the test results more universal and comparable. Therefore, the smart lock power consumption test method and system provided by the present invention overcome many defects in the prior art, not only improving the accuracy and efficiency of the test, but also providing technical tools for the design optimization, use management, and life prediction of smart locks, which is expected to promote the smart lock industry to develop in a more energy-saving and intelligent direction. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 This is a flow chart of a method for testing power consumption of a smart lock in one embodiment of the present invention; Figure 2 A flowchart of adaptively and dynamically adjusting the sampling frequency in one embodiment of the present invention; Figure 3 A line graph showing energy efficiency changes of a smart lock in one embodiment of the present invention; Figure 4 A flowchart for calculating the total power consumption of a smart lock within a preset time period in one embodiment of the present invention; Figure 5 A flowchart of dynamically adjusting the network wake-up interval in one embodiment of the present invention; Figure 6 This is an architecture diagram of a smart lock power consumption test system in one embodiment of the present invention; In the accompanying drawings, the components represented by the reference numerals are listed as follows: 100, log acquisition module; 200, state identification module; 300, power consumption acquisition module; 400, correlation model establishment module; 500, power consumption calculation module; 600, environment correction module. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] The term "comprise" and any variation thereof in the specification and claims of the present application are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. In addition, the use of "and / or" in the specification and claims to indicate at least one of the connected objects, such as A and / or B, means that A alone, B alone, and A and B exist in three cases.

[0012] In the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0013] The present invention provides the following preferred embodiments: Embodiment 1 In order to solve the problem of insufficient accuracy in the power consumption test of the smart lock, this embodiment proposes a test method for the power consumption test of the smart lock, so as to achieve a comprehensive, efficient and accurate test of the power consumption of the smart lock.

[0014] like Figure 1 As shown, the steps of the test method include: S101. Obtain the operation log of the smart lock within a preset time period.

[0015] S102. Identify the working status sequence of the smart lock based on the operation log.

[0016] S103: Collect real-time power consumption data corresponding to each working state, and dynamically and adaptively adjust the sampling frequency.

[0017] S104. Use a machine learning algorithm to establish a correlation model between the working state and power consumption of the smart lock.

[0018] S105. Calculate the total power consumption of the smart lock within a preset time period based on the association model.

[0019] S106: Acquire environmental factor data, and perform correction calculation to obtain total corrected power consumption.

[0020] First, the method obtains the operation log of the smart lock within a preset time period. This step can be achieved by connecting to the communication interface of the smart lock or reading log data from the internal storage of the smart lock. The operation log contains all operation records of the smart lock within the preset time period, which provide basic data for subsequent status identification.

[0021] Furthermore, based on the acquired operation log, the method identifies the working state sequence of the smart lock. The identification process can use a predefined state mapping table to map each operation record in the operation log to the corresponding working state. For example, the "fingerprint verification" operation can be mapped to the "identification state", and the "motor rotation" operation can be mapped to the "unlocking state", etc. In this way, the continuous operation log can be converted into a discrete working state sequence.

[0022] Furthermore, the method collects real-time power consumption data corresponding to each working state and adaptively and dynamically adjusts the sampling frequency. A high-precision current sensor is used in the collection process, and an analog-to-digital converter is used to convert the analog signal into a digital signal. The mechanism of adaptively and dynamically adjusting the sampling frequency can adjust the sampling frequency in real time according to the power consumption change rate, increase the sampling frequency when the power consumption changes dramatically, and reduce the sampling frequency when the power consumption is relatively stable, thereby optimizing the use of system resources while ensuring data accuracy.

[0023] It should be understood that this method uses a machine learning algorithm to establish a correlation model between the working state and power consumption of the smart lock. The machine learning algorithm used here can be a multi-layer perceptron neural network, whose input layer receives working state parameters and environmental factor data, the hidden layer uses the ReLU activation function for feature extraction and nonlinear mapping, and the output layer predicts the corresponding power consumption value. The model is trained through the back propagation algorithm and the Adam optimizer, and the model accuracy is continuously optimized through regular retraining.

[0024] Furthermore, the total power consumption of the smart lock within a preset time period is calculated based on the established association model. The calculation process first decomposes the working state sequence within the preset time period into multiple time segments, each time segment corresponds to a specific working state. Then, a feature vector is constructed for each time segment, the power consumption is predicted using the association model, and a time weighting factor is introduced for weighted calculation. Finally, the weighted power consumption of all time segments is accumulated to obtain the total power consumption within the preset time period.

[0025] Furthermore, the method obtains environmental factor data and uses it to correct the calculated total power consumption. The environmental factor data includes temperature, humidity, light intensity, and atmospheric pressure. The correction process uses a multiple linear regression model to consider the impact of various environmental factors on power consumption, thereby obtaining a more accurate total corrected power consumption.

[0026] It can be understood that there is a close logical relationship between the various steps in this method. The acquisition of operation logs provides a basis for state identification; the state identification results guide the collection of power consumption data; the collected power consumption data is used to train the association model; the association model is used to calculate the total power consumption; finally, the introduction of environmental factors further improves the accuracy of power consumption calculation.

[0027] The benefit of this embodiment is that, through adaptive sampling and machine learning modeling, the method improves the accuracy of power consumption testing. At the same time, the consideration of environmental factors makes the test results more universal. In addition, the establishment of the association model not only improves the test efficiency, but also provides an important basis for the design improvement and use management of smart locks, which helps to improve product performance.

[0028] Embodiment 2 In order to solve the contradiction between data accuracy and system energy consumption caused by the fixed sampling frequency during the power consumption test of the smart lock, this embodiment further optimizes the step of adaptively and dynamically adjusting the sampling frequency. This optimization scheme reduces the energy consumption of the test system while ensuring data accuracy by adjusting the sampling frequency in real time.

[0029] like Figure 2 As shown, the steps of adaptively and dynamically adjusting the sampling frequency include: S201 : Set an initial sampling frequency and a power consumption change threshold.

[0030] S202: Monitor the rate of change of power consumption data in real time.

[0031] S203: When the power consumption change rate exceeds a preset threshold, increase the sampling frequency.

[0032] S204: When the power consumption change rate is lower than a preset threshold, reduce the sampling frequency.

[0033] S205. By dynamically adjusting the sampling frequency, the energy consumption of the test system is reduced while ensuring data accuracy.

[0034] Specifically, the selection of the initial sampling frequency needs to take into account the general working characteristics of the smart lock, and can be set to about 1kHz, which is usually sufficient to capture the power consumption changes in most working states of the smart lock. The power consumption change threshold can be set between 5% and 10% to effectively balance sensitivity and stability.

[0035] Furthermore, the rate of change of power consumption data is monitored in real time. This step can be achieved by calculating the difference between two adjacent sampling data and dividing it by the time interval. In order to reduce the influence of noise, the sliding window average method can be used, that is, the average rate of change of the most recent N sampling points is calculated. The value of N can be adjusted according to the actual situation. Usually, a value between 5 and 10 can achieve good results.

[0036] It is important to understand that when the rate of change of power consumption exceeds a preset threshold, the system will increase the sampling frequency. The specific operation of increasing the sampling frequency can be multiplied by 2, that is, multiplying the current sampling frequency by 2. At the same time, in order to prevent the sampling frequency from being too high and causing the system to be overloaded, an upper limit value should be set, such as 10kHz. This ensures that more detailed information can be captured when the power consumption changes sharply.

[0037] Accordingly, when the power consumption change rate is lower than the preset threshold, the system will reduce the sampling frequency. The sampling frequency can be reduced by half, that is, the current sampling frequency is divided by 2. Similarly, in order to ensure the basic data collection quality, a lower limit value should be set, such as 100Hz. This mechanism can reduce unnecessary data collection when the power consumption is relatively stable, thereby reducing system energy consumption.

[0038] It is understandable that by dynamically adjusting the sampling frequency, this embodiment reduces the energy consumption of the test system while ensuring data accuracy. This adaptive mechanism can flexibly adjust the sampling strategy according to the actual situation of power consumption changes, avoiding the energy waste caused by fixed high-frequency sampling and overcoming the problem of insufficient data accuracy that may be caused by fixed low-frequency sampling.

[0039] Furthermore, in order to make the adaptive mechanism more robust, the concept of hysteresis interval can be introduced. That is, a buffer is set between the thresholds for increasing and decreasing the sampling frequency to prevent frequent switching of the sampling frequency when the power consumption fluctuates near the threshold. For example, the threshold for increasing the frequency can be set to 8%, and the threshold for decreasing the frequency can be set to 5%, so that the change rate between 5% and 8% will not trigger the frequency adjustment.

[0040] In addition, this embodiment may also consider introducing a prediction mechanism. By analyzing the change pattern of historical power consumption data, the power consumption change trend in the future period is predicted, so as to adjust the sampling frequency in advance. This forward-looking adjustment can further improve the response speed and efficiency of the system.

[0041] It should be noted that the adjustment of the sampling frequency should also take into account the limitations of the hardware. Different analog-to-digital converters and processors have their maximum supported sampling frequencies. Exceeding this frequency may cause data loss or system instability. Therefore, in the actual implementation process, the upper limit of the sampling frequency should be set according to the performance of the hardware used.

[0042] The benefit of this embodiment is that by adaptively and dynamically adjusting the sampling frequency, a balance is achieved between data accuracy and system energy consumption. This method not only improves the efficiency of power consumption testing, but also reduces the energy consumption of the test system itself, making it possible to continuously monitor the power consumption of smart locks for a long time. At the same time, the design idea of ​​this adaptive mechanism can also be extended to other scenarios that require long-term monitoring but are resource-constrained, and has broad practical value.

[0043] Embodiment 3 In order to solve the problem that the complex nonlinear relationship between the working state and power consumption in the power consumption test of the smart lock is difficult to accurately model, this embodiment further optimizes the selection and implementation method of the machine learning algorithm. Specifically, a multi-layer perceptron neural network is used to establish a correlation model between the working state and power consumption of the smart lock, and high-precision power consumption prediction is achieved through the network structure and training strategy.

[0044] In this embodiment, the machine learning algorithm is a multi-layer perceptron neural network, and the steps of the multi-layer perceptron neural network to establish a correlation model between the working state and power consumption of the smart lock include: Its input layer receives working status parameters and environmental factor data; its hidden layer uses the ReLU activation function for feature extraction and nonlinear mapping; its output layer predicts the corresponding power consumption value; it uses the back-propagation algorithm and Adam optimizer for model training; and it continuously optimizes the model accuracy through regular retraining.

[0045] Specifically, this embodiment constructs a multi-layer perceptron neural network, whose input layer receives working state parameters and environmental factor data. Working state parameters may include current operation type (such as standby, identification, unlocking, etc.), operation duration, motor speed, etc. Environmental factor data include temperature, humidity, light intensity, and atmospheric pressure, etc. These input parameters need to be normalized to ensure that data of different dimensions can be compared and calculated on the same scale. Normalization can use the minimum-maximum scaling method or the Z-score normalization method.

[0046] Furthermore, the hidden layer of the network uses the ReLU activation function for feature extraction and nonlinear mapping. The choice of the ReLU function is based on its excellent performance in deep learning. It can effectively alleviate the gradient vanishing problem, speed up training, and is simple and efficient in calculation. The number of hidden layers and the number of neurons in each layer need to be determined through experiments. Generally, you can start with 3-5 hidden layers, 100-200 neurons per layer, and then adjust according to the model performance. In order to prevent overfitting, you can add a Dropout layer between the hidden layers and set a dropout rate of 0.2-0.5.

[0047] It is important to understand that the output layer of the network predicts the corresponding power consumption value. The output layer uses a linear activation function to directly output the predicted power consumption value. Considering the characteristics of power consumption prediction, the mean square error (MSE) or mean absolute error (MAE) can be used as the loss function.

[0048] Furthermore, in terms of model training, this embodiment uses the back propagation algorithm and the Adam optimizer. The back propagation algorithm can efficiently calculate the gradients of all weights in the network, while the Adam optimizer combines the advantages of momentum and adaptive learning rate, and can quickly converge and adapt to the learning rates of different parameters. The initial learning rate can be set to 0.001, and then a learning rate decay strategy is used, such as halving the learning rate every 100 epochs.

[0049] It is understandable that in order to continuously optimize the model accuracy, this embodiment adopts a regular retraining strategy. It can be set to retrain once every certain period of time (such as every week or every month) or after accumulating a certain amount of new data. During the retraining process, the idea of ​​transfer learning can be adopted, and the previously trained model weights can be used as the initial value, which can speed up the convergence speed and maintain the stability of the model.

[0050] Furthermore, in order to improve the generalization ability of the model, the present embodiment may also introduce data enhancement technology. New training samples may be generated by adding a small amount of random noise, shifting or scaling the time series, etc. This not only increases the diversity of the training data, but also improves the model's ability to respond to small changes.

[0051] Furthermore, this embodiment also implements model interpretability analysis. By calculating the importance score of the input features (such as using SHAP values), the factors that have the greatest impact on power consumption prediction can be identified. This information not only helps to further optimize the model structure, but also provides valuable insights for the hardware design and software optimization of the smart lock.

[0052] It should be noted that the computational complexity and storage requirements of the model should be considered in actual deployment. Model compression techniques, such as weight quantization or knowledge distillation, can be used to reduce the model size and speed up the inference process, so that it can run in resource-constrained embedded systems.

[0053] The benefit of this embodiment is that the association model established by the multi-layer perceptron neural network can accurately capture the complex nonlinear relationship between the working state of the smart lock and the power consumption. Regular retraining and data enhancement strategies ensure that the model can adapt to changes in the use of the smart lock and different environmental conditions. This dynamic optimization modeling method not only improves the accuracy of power consumption prediction, but also provides reliable technical support for long-term performance monitoring and optimization of smart locks. At the same time, the interpretability analysis of the model provides a data-driven decision-making basis for product improvement, which helps to improve the energy efficiency of smart locks, such as Figure 3 As shown, the shaded area between the two broken lines is the energy efficiency improvement value of the smart lock.

[0054] Embodiment 4 This embodiment further optimizes the method for calculating the total power consumption of the smart lock within a preset time period based on the association model. The method calculates the power consumption of the smart lock through steps such as time segment decomposition, feature vector construction, and weighting factor introduction.

[0055] like Figure 4 As shown, the steps of calculating the total power consumption of the smart lock within a preset time period based on the association model include: S401: Decompose a sequence of working states within a preset time period into multiple time segments, each time segment corresponding to a specific working state.

[0056] S402: For each time segment, construct a feature vector including working state parameters and duration, where the feature vector is expressed as: , where s i Indicates the state type, t i Indicates duration, p i Indicates the working status parameter, e i Indicates environmental factors.

[0057] S403: Use the association model to predict the power consumption of each time segment. The prediction formula is expressed as: , where f represents the trained association model.

[0058] S404. According to the different durations of different states, a time weighting factor is introduced. The weighting factor calculation formula is: .

[0059] S405: The weighted power consumption of all time segments is accumulated to obtain the total power consumption in a preset time period. The total power consumption is calculated as follows: , where P i,j is the jth observation value of the i-th time segment, is the average power consumption of the time segment, and α is a positive coefficient.

[0060] In this embodiment, the working state sequence within the preset time period is decomposed into multiple time segments, each time segment corresponds to a specific working state. This decomposition method can accurately capture the working state changes of the smart lock at different time points, providing a fine-grained basis for subsequent power consumption calculation. The division of time segments can be based on fixed time intervals or state change triggers, and the specific selection should be based on the working characteristics of the smart lock.

[0061] Furthermore, for each time segment, a feature vector containing working state parameters and duration is constructed. This representation method encodes all relevant information of the time segment into a unified vector form, which is convenient for subsequent model processing.

[0062] It is important to understand that the power consumption of each time segment is predicted using the association model. This step inputs the feature vector constructed previously into the machine learning model to obtain the power consumption prediction value of each time segment. The association model function can be the multi-layer perceptron neural network mentioned above, or other suitable machine learning models, such as random forest or gradient boosting tree.

[0063] It is understandable that, since different states have different durations, this embodiment introduces a time weighting factor. The weighting method takes into account the contribution of each state to the total power consumption, making the calculation result more accurate.

[0064] Furthermore, the weighted power consumption of all time segments is accumulated to obtain the total power consumption within the preset time period. The calculation formula for the total power consumption has been recorded in the previous part of this embodiment. The formula not only takes into account the predicted power consumption of each time segment, but also takes into account the volatility of power consumption by introducing the variance term, so that the calculation result is more stable. It should be noted that the choice of the α coefficient has a certain influence on the calculation result. The larger the α value, the more attention is paid to the volatility of power consumption; the smaller the α value, the more emphasis is placed on the average power consumption. The optimal α value can be determined by methods such as cross-validation or grid search.

[0065] Furthermore, in order to improve the computing efficiency, parallel computing technology can be considered. Since the power consumption calculation of each time segment is relatively independent, these calculation tasks can be assigned to multiple processor cores and performed simultaneously, thereby significantly reducing the total computing time. In addition, in order to handle possible outliers, an anomaly detection mechanism can be introduced in the accumulation process. For example, a moving average or a statistical method can be used to identify power consumption values ​​that significantly deviate from the normal range, and these outliers can be appropriately processed, such as elimination or smoothing.

[0066] It is understandable that in practical applications, the real-time nature of the calculation also needs to be considered. For long-term power consumption calculations, a sliding window approach can be used to regularly update the total power consumption value, rather than waiting until the entire preset time period ends before performing a one-time calculation. This approach can provide near-real-time power consumption monitoring results.

[0067] The benefit of this embodiment is that, through fine-grained time segment decomposition and feature vector construction, combined with time weighting factors and power consumption fluctuations, accurate calculation of the power consumption of the smart lock during long-term operation is achieved. This method not only considers the duration differences of different working states, but also captures the dynamic change characteristics of power consumption by introducing variance terms.

[0068] Embodiment 5 In order to solve the impact of environmental factors on the test results during the power consumption test of the smart lock, this embodiment further optimizes the method of correcting the total power consumption. This method introduces a multivariate linear regression model, considers environmental factors such as temperature, humidity, light intensity and atmospheric pressure, and accurately corrects the total power consumption, thereby improving the accuracy and reliability of the power consumption test.

[0069] In this embodiment, the environmental factor data includes temperature, humidity, light intensity and atmospheric pressure, and the steps of correcting and calculating to obtain the total corrected power consumption include: Get data on temperature, humidity, light intensity, and atmospheric pressure; The total power consumption is corrected using a multivariate linear regression model, and the correction formula is: , where P corrected is the total corrected power consumption, T, H, L, and P are the normalized temperature, humidity, light intensity, and atmospheric pressure, respectively, and β, γ, δ, and ε are the influence coefficients, respectively.

[0070] Specifically, obtain data on temperature, humidity, light intensity, and atmospheric pressure. The collection of data on these environmental factors requires the use of high-precision sensors to ensure the accuracy of the data. The temperature sensor can use a PT100 platinum resistor or a thermocouple, with a measurement range of -40°C to 85°C and an accuracy of ±0.1°C. The humidity sensor can use a capacitive humidity sensor with a measurement range of 0-100%RH and an accuracy of ±2%RH. The light intensity sensor can use a silicon photocell or a photoresistor with a measurement range of 0-100000lux and an accuracy of ±5%. The atmospheric pressure sensor can use a MEMS pressure sensor with a measurement range of 300-1100hPa and an accuracy of ±0.1hPa. These sensors should be integrated with the smart lock power consumption test system to achieve synchronous data acquisition.

[0071] Furthermore, the collected environmental factor data are preprocessed and normalized. Preprocessing includes removing outliers and smoothing. Outliers can be identified and eliminated by the 3σ criterion or the interquartile range method. Smoothing can be performed by moving average or low-pass filtering to reduce random fluctuations in the data. Normalization can be performed by the minimum-maximum scaling method or the Z-score standardization method to unify environmental factor data of different dimensions to the same scale, which is convenient for subsequent multivariate linear regression analysis.

[0072] It should be understood that this embodiment uses a multivariate linear regression model to correct the total power consumption. The correction formula assumes that the impact of environmental factors on power consumption is linearly superimposed, and the impact degree of different environmental factors on power consumption can be reflected by adjusting the impact coefficient.

[0073] It is understandable that the coefficients β, γ, δ, and ε can be obtained by regression analysis of a large amount of historical data. The specific steps include: collecting enough power consumption data and corresponding environmental factor data; fitting the multivariate linear model using the least squares method or other regression algorithms; evaluating the significance of each coefficient through the t-test or F-test; and using the determination coefficient R² and the root mean square error RMSE to evaluate the fitting effect of the model. In order to improve the stability and generalization ability of the model, resampling techniques such as cross-validation or bootstrap can be used.

[0074] Furthermore, in order to deal with possible nonlinear relationships, nonlinear transformations can be considered. For example, using a quadratic term T² for temperature or a logarithmic transformation log(L) for light intensity. These nonlinear transformations can be determined by observing scatter plots or using generalized additive models (GAMs).

[0075] The benefit of this embodiment is that the total power consumption is corrected for environmental factors through a multivariate linear regression model, effectively reducing the interference of external environmental changes on the power consumption test results. This method not only takes into account the combined impact of multiple environmental factors, but also determines the impact weight of each factor in a data-driven manner, making the correction results more objective and accurate. Through the correction method of this embodiment, comparable power consumption data can be obtained under different environmental conditions, providing a more reliable basis for the performance evaluation and optimization of smart locks.

[0076] Embodiment 6 In order to solve the problem of high standby power consumption when the smart lock is connected to the network, this embodiment further optimizes the network connection management strategy. This method implements an adaptive network connection strategy by monitoring the impact of the network connection status on power consumption, thereby effectively reducing standby power consumption while ensuring the normal communication function of the smart lock.

[0077] In this embodiment, the steps of the testing method further include: Monitor the impact of the smart lock's network connection status on power consumption; dynamically adjust the network wake-up interval based on usage frequency and importance through adaptive network connection strategies; achieve a balance between network connection and power consumption to reduce standby power consumption.

[0078] Specifically, this embodiment uses a high-precision current sampling chip, such as INA226 or MAX4080, to sample the power supply current of the smart lock in real time, and the sampling frequency can be set to 1kHz. At the same time, the network status detection module is used to record the network activities of the smart lock, including events such as connection establishment, data transmission, and connection disconnection. By performing time synchronization analysis on the current sampling data and the network activity log, the power consumption characteristics under different network states can be accurately quantified.

[0079] Furthermore, based on the monitoring data, this embodiment constructs a relationship model between network connection status and power consumption. Time series analysis methods, such as autoregressive integrated moving average model (ARIMA) or long short-term memory network (LSTM), are used to model the power consumption fluctuation caused by changes in network connection status. This dynamic model can capture the short-term and long-term impact of network connection on power consumption, providing a basis for subsequent adaptive strategies.

[0080] It should be understood that the adaptive network connection strategy in this embodiment dynamically adjusts the network wake-up interval according to the usage frequency and importance. The specific implementation method is as follows: First, define a usage frequency index F, and the solution formula of the usage frequency index F is as follows: , where N C is the actual number of connections, N I is the number of important connections (such as remote unlocking, security alarm, etc.), w is the importance weight, and T is the observation time window. Next, an adaptive algorithm is set to dynamically adjust the network wake-up interval τ according to the value of F. The solution formula for the network wake-up interval τ is as follows: , where τ min and τ max are the minimum and maximum wake-up intervals, respectively, and k is the adjustment coefficient. The formula ensures that the wake-up interval is close to τ when the usage frequency is high min , to ensure the response speed; when the usage frequency is low, the wake-up interval is gradually increased to τ max , reducing power consumption.

[0081] It is understandable that in order to achieve a balance between network connection and power consumption, this embodiment introduces a multi-level sleep mechanism. Three states are defined: light sleep, moderate sleep, and deep sleep, corresponding to different power consumption levels and wake-up delays. In the light sleep state, some functions of the network module are maintained, and connection requests can be responded to quickly; in the deep sleep state, the network module is completely turned off, and only the most basic wake-up circuit is retained. The system dynamically selects the appropriate sleep level based on the wake-up interval τ calculated by the adaptive algorithm.

[0082] Furthermore, this embodiment implements an intelligent pre-wake-up mechanism. By analyzing historical usage patterns, possible high-frequency usage periods are predicted. Before these periods, the wake-up interval is reduced in advance to ensure user experience. The prediction model can use a time series prediction method, such as the Prophet algorithm or a recurrent neural network (RNN).

[0083] The benefit of this embodiment is that by accurately monitoring the impact of network connection status on power consumption, combined with adaptive network connection strategy and multi-level sleep mechanism, a dynamic balance between smart lock network connection and power consumption is achieved. This method not only significantly reduces the standby power consumption of the smart lock, but also ensures the network response performance at critical moments.

[0084] Embodiment 7 In order to solve the contradiction between power consumption and response speed in network connection management of smart locks, this embodiment further optimizes the adaptive network connection strategy. This method sets the network connection importance threshold and power consumption budget, combines historical usage data prediction, and realizes dynamic adjustment of the network wake-up interval, thereby minimizing network power consumption while ensuring the response speed of the smart lock.

[0085] like Figure 5 As shown, the steps of dynamically adjusting the network wake-up interval according to the usage frequency and importance through the adaptive network connection strategy include: S701: Set a network connection importance threshold and a power consumption budget.

[0086] S702: Predict the time of the next network connection requirement based on historical usage data.

[0087] S703: Increase the frequency of network inspection near the predicted time point.

[0088] S704: Extend the network sleep time during non-critical periods.

[0089] S705. By dynamically adjusting the strategy, the network power consumption is minimized while ensuring the response speed.

[0090] Specifically, the network connection importance threshold can be represented by defining a multidimensional vector, including factors such as connection type, data transmission volume, and security level. For example, the importance of a remote unlocking command can be set to 0.9, while ordinary state synchronization can be set to 0.3. The power consumption budget is determined based on the battery capacity and expected usage cycle of the smart lock, and can be expressed as a daily average power consumption limit, such as 50mAh / day. The setting of these parameters provides a decision basis for subsequent dynamic adjustment strategies.

[0091] Furthermore, this embodiment develops a network connection demand prediction model based on historical usage data. The network connection mode of the smart lock is modeled using time series analysis methods, such as the autoregressive integrated moving average model (ARIMA) or the long short-term memory network (LSTM). The model input includes features such as historical connection timestamps, connection types, and durations, and the output is the predicted time and type of the next network connection. In order to improve the prediction accuracy, the model also considers periodic factors (such as weekday / weekend patterns) and external factors (such as weather conditions, user schedules, etc.). The prediction model is updated regularly (such as weekly) with the latest data to adapt to changes in user habits.

[0092] It should be understood that, based on the prediction results, this embodiment implements dynamic adjustment of the network check frequency. Near the predicted network connection time point, the system will increase the network check frequency. Ensure that the network check frequency reaches a peak near the predicted time point and then gradually decreases, thereby ensuring response speed while avoiding unnecessary power consumption.

[0093] It is understandable that in non-critical periods, this embodiment adopts a strategy of extending the network sleep time, ensuring that during continuous inactivity, the sleep time gradually increases but does not exceed a preset maximum value, thereby balancing power saving and response speed.

[0094] Furthermore, this embodiment implements a dynamic adjustment strategy to minimize network power consumption while ensuring response speed. The system continuously optimizes this cost function by adjusting the network check frequency and sleep time. Optimization methods such as gradient descent or genetic algorithm are used to find the optimal parameter combination under the constraints of the power consumption budget. It should be noted that in order to handle sudden high-priority network needs, this embodiment also designs an emergency wake-up mechanism. When a specific high-priority signal (such as a security alarm) is received, the system will immediately exit the sleep state and return to the highest network check frequency. This mechanism is implemented through hardware interrupts to ensure a rapid response in the event of an emergency.

[0095] Furthermore, this embodiment also introduces a self-learning mechanism, which dynamically adjusts the importance threshold and the parameters of the prediction model by recording the user's feedback on the network response (such as the number of repeated connection attempts). This self-learning capability enables the system to gradually adapt to the usage habits of different users and provide personalized network connection management strategies.

[0096] The benefit of this embodiment is that by comprehensively considering the importance of network connection, power consumption budget, historical usage pattern and real-time feedback, the refined control of smart lock network connection management is achieved. It not only reduces the standby power consumption of the smart lock, but also ensures the rapid response capability at critical moments.

[0097] Embodiment 8 In order to solve the problem of predicting the service life of smart locks, this embodiment further optimizes the detection method of hardware aging and the estimation model of the remaining service life. This method detects the hardware aging of the smart lock and establishes a prediction model of hardware aging and power consumption growth, thereby accurately estimating the remaining service life of the smart lock and providing users with timely maintenance and replacement suggestions.

[0098] In this embodiment, the steps of the testing method further include: Detect the hardware aging degree of smart locks; Establish a prediction model for hardware aging and power consumption growth; Estimate the remaining service life of the smart lock based on the prediction model; The calculation formula of the prediction model is: , where L is the remaining service life, L0 is the initial design life, k is the attenuation coefficient, P threshold is the power consumption threshold, P current is the current power consumption.

[0099] Specifically, this embodiment realizes a comprehensive detection of the aging degree of the smart lock hardware. A multi-parameter monitoring method is adopted, including but not limited to the following aspects: power management unit (PMU) efficiency monitoring, microcontroller (MCU) performance degradation evaluation, sensor sensitivity change detection, and mechanical parts wear degree measurement. For PMU efficiency monitoring, a high-precision current sampling chip (such as TI's INA226) is used to monitor the input and output currents in real time and calculate the change trend of the conversion efficiency. The MCU performance degradation evaluation is performed by regularly executing standardized benchmark programs and recording changes in execution time and power consumption. Sensor sensitivity change detection is achieved by comparing the deviation of the sensor output from the standard calibration value. The degree of wear of mechanical parts can be evaluated by measuring the change in motor drive current or using an acceleration sensor to detect vibration characteristics during the switching process.

[0100] Furthermore, this embodiment establishes a prediction model for hardware aging and power consumption growth. The model uses a multivariate nonlinear regression method, taking the above-described hardware aging indicators as input variables and the power consumption growth rate as output variables.

[0101] It should be understood that in order to improve the accuracy and adaptability of the model, this embodiment introduces an adaptive learning mechanism. The system regularly collects actual power consumption data and various aging indicators and compares them with the model prediction results. If the prediction error exceeds the preset threshold, the automatic adjustment of the model parameters is triggered. The adjustment process uses an online learning algorithm, such as the stochastic gradient descent method, to minimize the prediction error. This dynamic adjustment mechanism enables the model to adapt to the hardware aging characteristics in different usage environments and modes.

[0102] It can be understood that based on the above prediction model, this embodiment develops a method for estimating the remaining service life of the smart lock. The estimation formula assumes that there is an exponential relationship between power consumption growth and life reduction, which can better reflect the nonlinear effect of hardware aging on life. The attenuation coefficient k reflects the degree of influence of power consumption growth on life, and its optimal value can be determined by analyzing a large amount of historical data.

[0103] Furthermore, this embodiment also implements the confidence interval calculation of life prediction. The Monte Carlo simulation method is used to consider the uncertainty of various input parameters, generate a large amount of simulation data, and calculate the probability distribution of life prediction. This can not only provide a point estimate, but also provide a 95% confidence interval, providing users with more comprehensive life information.

[0104] It should be noted that this embodiment has designed a graded warning mechanism. According to the estimated remaining service life, the warning is divided into three levels: mild, moderate and severe. When entering the mild warning stage, the system will recommend the user to perform simple maintenance; in the moderate warning stage, it will recommend in-depth inspection and component replacement; in the severe warning stage, it is recommended to replace the entire device. This graded warning mechanism helps users to arrange maintenance plans reasonably and avoid the inconvenience caused by sudden equipment failure.

[0105] The benefit of this embodiment is that by comprehensively detecting the degree of hardware aging, combined with an adaptive prediction model and environmental factor correction, an accurate estimation of the remaining service life of the smart lock is achieved. This method not only takes into account the impact of hardware aging on power consumption, but also takes into account the interference of environmental factors, thereby providing a more reliable life prediction result. Through the prediction model and early warning mechanism of this embodiment, users can timely understand the status of the device, reasonably arrange maintenance and replacement plans, and improve the safety of smart locks.

[0106] Embodiment 9 In order to solve the problems of data collection, state identification, environmental impact, etc. in the process of power consumption test of smart locks, this embodiment proposes a power consumption test system for smart locks. The system realizes full-process automated testing from log acquisition to power consumption calculation through modular design, and introduces an environmental factor correction mechanism to improve the accuracy and reliability of power consumption testing.

[0107] like Figure 6 As shown, the test system includes: The log acquisition module 100 is used to obtain the operation log of the smart lock within a preset time period.

[0108] The state identification module 200 is used to identify the working state sequence of the smart lock based on the operation log.

[0109] The power consumption collection module 300 is used to collect real-time power consumption data corresponding to each working state according to a dynamically adjusted sampling frequency.

[0110] The association model building module 400 is used to establish an association model between the working state and power consumption of the smart lock using a machine learning algorithm.

[0111] The power consumption calculation module 500 is used to calculate the total power consumption of the smart lock within a preset time period based on the association model.

[0112] The environment correction module 600 is used to obtain environmental factor data and correct the calculated total power consumption.

[0113] In this embodiment, a log acquisition module 100 is provided. The log acquisition module 100 adopts a non-intrusive data acquisition method to acquire the operation log in real time through the debugging interface or the wireless communication interface of the smart lock. The log data includes but is not limited to: unlocking / locking operation timestamps, network connection status changes, motor drive records, sensor trigger events, etc. In order to ensure the integrity and real-time nature of the data, a double buffer technology is used for log storage, and a CRC check is used to ensure the accuracy of data transmission. The log acquisition module 100 also implements a data compression function, using a combination of differential coding and Huffman coding to effectively reduce the overhead of data transmission and storage.

[0114] Furthermore, the state recognition module 200 uses a combination of a finite state machine (FSM) and a hidden Markov model (HMM) to identify the working state sequence of the smart lock based on the acquired operation log. FSM is used to process clear state transitions, such as unlocking, locking, network connection, etc.; HMM is used to process states with higher uncertainty, such as standby, light sleep, deep sleep, etc. The state recognition algorithm uses sliding window technology to process the input log data stream in real time to ensure the timeliness of state recognition. At the same time, an adaptive learning mechanism is introduced to continuously optimize the state transition probability matrix and the emission probability matrix by comparing the accumulated recognition results with the actual power consumption data, thereby improving the accuracy of recognition.

[0115] It should be understood that the power consumption acquisition module 300 adopts a dynamically adjusted sampling strategy. Based on the INA226 high-precision current monitoring chip of Texas Instruments, a wide range (from μA to A level) of power consumption measurement is achieved. This strategy ensures that a higher sampling frequency is used before and after the state transition to capture transient power consumption changes, while the sampling frequency is reduced in the stable state to reduce the data processing burden.

[0116] It is understandable that the association model building module 400 uses a machine learning algorithm to construct an association model between the working state and power consumption of the smart lock. Taking into account the nonlinearity and time-varying nature of the power consumption characteristics of the smart lock, this embodiment adopts a deep learning method, specifically a hybrid model of a long short-term memory network (LSTM) and a convolutional neural network (CNN). LSTM is used to capture the long-term dependence of power consumption, while CNN is used to extract local features. The model input includes working state sequences, environmental parameters, historical power consumption data, etc., and the output is the predicted power consumption value. The model training uses the Adam optimizer and uses the root mean square error (RMSE) as the loss function. In order to prevent overfitting, L2 regularization and Dropout technology are introduced. In addition, an online learning mechanism is implemented to regularly update the model parameters with newly collected data to ensure that the model can adapt to changes in the usage mode and hardware characteristics of the smart lock.

[0117] Furthermore, the power consumption calculation module 500 calculates the total power consumption of the smart lock within a preset time period based on the established association model and the identified working state sequence. The calculation process adopts the Monte Carlo integration method, taking into account the uncertainty of the state duration. Specifically, for each identified state, a large number of random duration samples are generated, the corresponding power consumption is calculated, and then the average is taken as the contribution of the state. This method not only gives a point estimate of the total power consumption, but also provides a confidence interval, providing more comprehensive information for power consumption evaluation.

[0118] It should be noted that the environment correction module 600 corrects the calculated total power consumption by acquiring environmental factor data. Environmental factors include temperature, humidity, atmospheric pressure, etc., which are collected in real time through an integrated environmental sensor module (such as Bosch BME280).

[0119] In addition, this embodiment also designs an anomaly detection module to identify abnormal patterns in power consumption data. This module uses an anomaly detection algorithm based on an autoencoder to detect abnormal power consumption behavior caused by hardware failure, software error, or external interference. Once an anomaly is detected, the system generates an alarm and marks the relevant data to ensure the reliability of the test results.

[0120] The benefit of this embodiment is that by integrating modules such as log acquisition, state recognition, power consumption collection, association model establishment, power consumption calculation and environmental correction, the power consumption test of the smart lock is fully automated. The dynamic sampling strategy, deep learning model and environmental correction algorithm adopted significantly improve the accuracy and adaptability of the power consumption test. Through the test system of this embodiment, the power consumption characteristics of the smart lock under various working states and environmental conditions can be deeply analyzed.

[0121] Embodiment 10 In order to solve the portability and scalability problems of the smart lock power consumption test method in practical applications, this embodiment further optimizes the design and implementation of the computer-readable storage medium. The computer program stored on the storage medium can execute the smart lock power consumption test method on different computing platforms, thereby improving the applicability and portability of the test system.

[0122] This embodiment adopts a modular software architecture design to encapsulate the various functional modules of the power consumption test method into independent program components. These components include log parser, state recognition engine, power consumption acquisition driver, data processing library, machine learning model, environmental factor correction algorithm, etc. Each component defines a standardized interface to facilitate system integration and future function expansion. The program is written in a cross-platform programming language (such as C++ or Python), and uses cross-platform build tools such as CMake to ensure that the code can be compiled and run on different operating systems such as Windows, Linux, and macOS.

[0123] Furthermore, this embodiment implements a parameter management mechanism driven by configuration files. Test parameters, device characteristics, environmental thresholds, etc. can be specified through external configuration files in JSON or YAML format, and can be adapted to different models of smart locks or test scenarios without modifying the source code. The program will automatically load and parse these configuration files at startup, and dynamically adjust internal parameters and algorithm behavior.

[0124] It should be understood that in order to improve the execution efficiency of the program, this embodiment adopts multi-threaded parallel processing technology. Time-consuming operations such as log acquisition, state identification, and power consumption collection are assigned to different threads for execution, and data exchange and synchronization are achieved through message queues between threads. This design fully utilizes the computing power of multi-core processors and significantly improves the real-time performance of the test system.

[0125] The benefit of this embodiment is that by designing a cross-platform, configurable, and modular software architecture, the high flexibility and scalability of the smart lock power consumption test method is achieved. Through the computer-readable storage medium of this embodiment, testers can easily deploy and execute power consumption tests in different computing environments, while also providing convenient framework support for future function upgrades and algorithm optimization.

[0126] The above embodiments further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A test method for testing the power consumption of a smart lock, characterized in that: The steps of the test method include: Obtaining the operation log of the smart lock within a preset time period; Identify the working state sequence of the smart lock based on the operation log; Collect real-time power consumption data corresponding to each working state and dynamically adjust the sampling frequency adaptively; Use machine learning algorithms to establish a correlation model between the working status and power consumption of smart locks; Calculate the total power consumption of the smart lock within the preset time period based on the association model; The environmental factor data is obtained and the total corrected power consumption is obtained by correction calculation.

2. The testing method according to claim 1, characterized in that: The step of adaptively and dynamically adjusting the sampling frequency comprises: Set the initial sampling frequency and power consumption change threshold; Real-time monitoring of the rate of change of power consumption data; When the power consumption change rate exceeds a preset threshold, the sampling frequency is increased; When the power consumption change rate is lower than a preset threshold, the sampling frequency is reduced; By dynamically adjusting the sampling frequency, the energy consumption of the test system can be reduced while ensuring data accuracy.

3. The testing method according to claim 1, characterized in that: The machine learning algorithm is a multi-layer perceptron neural network, and the steps of establishing a correlation model between the working state and power consumption of the smart lock by the multi-layer perceptron neural network include: Its input layer receives the working state parameters and the environmental factor data; Its hidden layer uses the ReLU activation function for feature extraction and nonlinear mapping; Its output layer predicts the corresponding power consumption value; Use back-propagation algorithm and Adam optimizer for model training; Continuously optimize model accuracy through regular retraining.

4. The testing method according to claim 1, characterized in that: The step of calculating the total power consumption of the smart lock within the preset time period based on the association model includes: Decomposing the working state sequence within a preset time period into a plurality of time segments, each of the time segments corresponding to a specific working state; For each of the time segments, a feature vector including working state parameters and duration is constructed, and the feature vector is expressed as: , where s i Indicates the state type, t i Indicates duration, p i Indicates the working status parameter, e i Indicates environmental factors; The power consumption of each time segment is predicted using the association model, and the prediction formula is expressed as: , where f represents the trained association model; According to the different durations of different states, a time weighting factor is introduced, and the calculation formula of the weighting factor is: ； The weighted power consumption of all the time segments is accumulated to obtain the total power consumption in the preset time period. The calculation formula of the total power consumption is: , where P i,j is the jth observation value of the ith time segment, is the average power consumption of the time segment, and α is a positive coefficient.

5. The testing method according to claim 4, characterized in that: The environmental factor data include temperature, humidity, light intensity and atmospheric pressure, and the step of correcting and calculating to obtain the total corrected power consumption includes: Acquiring data of the temperature, the humidity, the light intensity and the atmospheric pressure; The total power consumption is corrected using a multivariate linear regression model, and the correction formula is: , where P corrected is the total corrected power consumption, T, H, L, and P are the normalized temperature, humidity, light intensity, and atmospheric pressure, respectively, and β, γ, δ, and ε are the influence coefficients, respectively.

6. The testing method according to claim 1, characterized in that: The steps of the test method also include: Monitoring the impact of the network connection status of the smart lock on power consumption; Through adaptive network connection strategy, the network wake-up interval is dynamically adjusted according to the frequency and importance of use; Achieve a balance between network connectivity and power consumption to reduce standby power consumption.

7. The testing method according to claim 6, characterized in that: The step of dynamically adjusting the network wake-up interval according to the usage frequency and importance through the adaptive network connection strategy includes: Set network connection importance thresholds and power consumption budgets; Predict when the next network connection will be needed based on historical usage data; Increase the frequency of network checks near the prediction time point; Extend network sleep time during non-critical periods; By dynamically adjusting strategies, network power consumption can be minimized while ensuring response speed.

8. The testing method according to claim 1, characterized in that: The steps of the test method also include: Detecting the hardware aging degree of the smart lock; Establishing a prediction model for the hardware aging and power consumption growth; Based on the prediction model, estimating the remaining service life of the smart lock; The calculation formula of the prediction model is: , where L is the remaining service life, L0 is the initial design life, k is the attenuation coefficient, P threshold is the power consumption threshold, P current is the current power consumption.

9. A test system for testing the power consumption of a smart lock, characterized in that: The test system comprises: The log acquisition module is used to obtain the operation log of the smart lock within a preset time period; A state recognition module, used to identify the working state sequence of the smart lock based on the operation log; The power consumption collection module is used to collect the real-time power consumption data corresponding to each working state according to the dynamically adjusted sampling frequency; A correlation model building module is used to build a correlation model between the working state and power consumption of the smart lock using a machine learning algorithm; A power consumption calculation module, used to calculate the total power consumption of the smart lock within the preset time period based on the association model; The environmental correction module is used to obtain environmental factor data and correct the calculated total power consumption.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor of a computer, the testing method according to any one of claims 1 to 8 is implemented.

Images