A power control method and system suitable for a hybrid excavator

Abstract

The application discloses a power control method and system suitable for a hybrid excavator, and relates to the technical field of engineering machinery. The method comprises the following steps: collecting and preprocessing time series data; inputting the preprocessed time series data into a trained LSTM neural network model to predict a swing motor power prediction value at a future moment; predicting battery power according to the swing motor power prediction value, calculating the final demand power of an ISG according to the battery power, and calculating the target speed reduction of an electric swing mechanism according to whether the battery prediction power exceeds a power limit value; and converting the final demand power of the ISG and the target speed reduction of the electric swing mechanism into corresponding ISG motor and electric swing mechanism control signals. The method can predict power demand in advance through an LSTM network, actively stabilize the charging and discharging power and SOC of the battery in an ideal interval, effectively suppress power fluctuation, and prolong the service life of the battery.

Description

Technical Field

[0001] This invention relates to the field of new energy construction machinery technology, and in particular to a power control method and system suitable for hybrid excavators. Background Technology

[0002] Against the backdrop of energy conservation and emission reduction, the construction machinery industry is accelerating its green transformation. New energy technologies, due to their advantages such as low operating costs, energy saving and emission reduction, and powerful performance, are being widely applied in construction machinery products. In the electrification upgrade, the slewing mechanisms of construction machinery with slewing platforms, such as excavators and cranes, are being replaced from traditional hydraulic slewing mechanisms to electric slewing mechanisms. The electric slewing mechanism consists of a slewing motor and a reducer, driving the platform to rotate and recovering braking energy through electric braking, thereby improving the overall energy efficiency of the machine.

[0003] However, the electric slewing mechanism requires high bus power during start-up and shutdown, and frequent fluctuations in its power amplitude can have numerous negative impacts on the power battery, such as accelerating battery aging, increasing heat generation, shortening lifespan, and reducing the accuracy of state estimation. Drastic fluctuations in charge and discharge levels also accelerate the depletion of battery cycle life. Power batteries are expensive, and premature replacement could result in the overall operating cost of a hybrid excavator being no lower than that of a traditional gasoline-powered model. Therefore, smoothly controlling the battery's charge and discharge power and SOC (State of Charge) is of significant economic importance for reducing the operating cost of hybrid excavators and extending battery life.

[0004] Existing technologies (such as CN202510768734.2 and CN202510617471.5) mainly use ISG (Integrated Starter Generator) motors for peak shaving and valley filling. Under heavy loads, they assist the engine in outputting power, and under light loads, they generate electricity to charge the battery, thereby optimizing the engine's operating point or meeting the vehicle's power requirements. However, these solutions generally overlook the negative impact of fluctuations in charging and discharging power or SOC on battery life.

[0005] Therefore, there is an urgent need for a power control scheme that can effectively suppress battery power and SOC fluctuations in order to extend battery life. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a power control method and system suitable for hybrid excavators. This method and system can effectively predict and smooth the charging and discharging power of the power battery in the hybrid excavator and stabilize its SOC within the ideal range, thereby extending the battery life and reducing the overall operating cost of the machine.

[0007] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0008] In a first aspect, the present invention provides a power control method suitable for hybrid excavators, the method comprising:

[0009] Collect and preprocess time series data;

[0010] The preprocessed time series data is input into a trained LSTM neural network model to predict the power of the rotary motor at a future time. The input of the LSTM neural network model is multivariate time series data within a past time window, and the output is the predicted power of the rotary motor at a specified future time.

[0011] The battery power is predicted based on the predicted power value of the rotary motor, the final power requirement of the ISG is calculated based on the battery power, and the target speed reduction of the electric rotary mechanism is calculated based on whether the predicted battery power exceeds the power limit.

[0012] Based on the final power requirement of the ISG and the reduction in the target speed of the electric rotary mechanism, the corresponding control signals for the ISG motor and electric rotary mechanism are converted.

[0013] Furthermore, the battery power is predicted based on the predicted power value of the rotary motor, the final power requirement of the ISG is calculated based on the battery power, and the target speed reduction of the electric rotary mechanism is calculated based on whether the predicted battery power exceeds the power limit, including:

[0014] Calculate the first predicted power of the battery based on the predicted power value of the rotary motor;

[0015] Obtain the power limit of the ISG motor;

[0016] Based on the first predicted power of the battery, calculate the first target power increment of the ISG and the second target power increment of the ISG.

[0017] Calculate the final power requirement of the ISG based on the power limit of the ISG motor, the first target power increment of the ISG, and the second target power increment of the ISG.

[0018] Update the second predicted power of the battery based on the final power demand of the ISG;

[0019] The reduction in the target speed of the rotary motor is calculated based on the second predicted power of the battery.

[0020] Furthermore, the time series data includes time series data of handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current;

[0021] The collection and preprocessing of time series data includes:

[0022] The raw signals of handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current are collected by sensors.

[0023] The original signal is then subjected to timestamp alignment, filtering, and normalization.

[0024] Furthermore, the training method for the LSTM neural network model includes:

[0025] Construct a training dataset, wherein the input samples of the dataset are the time series data within a past time window, and the output samples are the actual rotary motor power values ​​at a specified future time.

[0026] Using mean squared error as the loss function, Adam is selected to optimize the parameter matrix and bias of each layer of the network, and the constructed LSTM neural network model is trained.

[0027] Furthermore, the LSTM neural network model includes an input layer, a first LSTM layer, a second LSTM layer, and an output layer connected in sequence;

[0028] The input layer is used to receive input data of shape (Ts, F), where Ts is the time step and F is the feature dimension;

[0029] The first LSTM layer is used to extract the temporal features of the input data, and its memory unit count is 64.

[0030] The second LSTM layer is used to further analyze the temporal features and output the feature vector of the final time step, and its memory unit number is 32;

[0031] The output layer is a fully connected layer used to map the feature vector into predicted values ​​of rotary motor power.

[0032] Furthermore, the first predicted power of the battery is calculated based on the predicted power value of the rotary motor, including:

[0033] ;

[0034] in, The first predicted power of the battery, For the predicted rotary motor power, This represents the current power of the ISG motor.

[0035] Furthermore, the power limits of the ISG motor are obtained, including:

[0036] Based on the battery's SOC and temperature, and referring to the battery's charge / discharge power MAP, the ideal charging power limit of the battery is calculated. Ideal discharge power limit Maximum charging power limit Maximum discharge power limit The charging power is negative, and the discharging power is positive.

[0037] The maximum available power of the ISG motor is calculated based on the hydraulic system power and engine power limits. Minimum available power The maximum available power is a positive value, indicating that the power is driving; the minimum available power is a negative value, indicating that the power is generating.

[0038] Calculate the minimum available power of the ISG motor during normal operation. The maximum available power is ;

[0039] in, This is the minimum power of the ISG motor (negative value when generating electricity). The maximum power of the ISG motor (positive when driven) and these two values ​​depend on the external characteristics of the motor.

[0040] This represents the current power of the hydraulic system, calculated by measuring the pressure and flow rate of the hydraulic pump, and is a positive power.

[0041] This represents the current engine power, and its power range is... This is related to the external characteristics of the engine.

[0042] Furthermore, the calculation of the first target power increment of the ISG includes:

[0043] Based on the first predicted power of the battery With respect to the ideal charging power limit of the battery and ideal discharge power limit The relationship is used to calculate the first power increment. The relationship is as follows:

[0044] when When the ideal charging power limit is between the ideal discharging power limit and the ideal charging power limit, i.e. ISG motor target power increment The ISG motor maintains its current power output;

[0045] when When the ideal discharge power limit is exceeded, i.e. ISG motor target power increment , When the value is negative, the ISG motor reduces its current power output;

[0046] when When the charging power is below the ideal charging power limit, i.e. ISG motor target power increment , When the value is positive, the ISG motor increases its current power output.

[0047] Furthermore, the calculation of the second target power increment of the ISG includes:

[0048] Based on the difference between the battery's current SOC and the target SOC, a preset mapping table is consulted to obtain the second power increment. .

[0049] The target Soc is 60%. The values ​​are obtained by looking up a table, as shown in Table 1, with the unit being Kw:

[0050] Table 1 Mapping Table of ISG Second Target Power Increment and Second Power Increment

[0051]

[0052] Furthermore, based on the power limit of the ISG motor, the first target power increment of the ISG, and the second target power increment of the ISG, the final power requirement of the ISG is calculated, including:

[0053] Based on the first target power increment of the ISG motor Second target power increment Calculate the final power demand of the ISG :

[0054] ;

[0055] Based on the minimum available power of the ISG motor Maximum available power Limit this value: reduce the final power demand. The value is limited to [ , ].

[0056] If the ISG's final power requirement Greater than the maximum available power of the ISG motor Then determine the final power requirement of the ISG. The value is the maximum available power of the ISG motor. ;

[0057] If the ISG's final power requirement Less than the minimum available power of the ISG motor Then determine the final power requirement of the ISG. The value is the minimum available power of the ISG motor. .

[0058] Furthermore, based on the ISG's final power demand, the second predicted power of the battery is updated, including:

[0059] The second predicted power of the battery is updated based on the final power demand of the ISG. :

[0060] Among them, the final required power , This represents the predicted power of the rotary motor.

[0061] Furthermore, based on the battery's second predicted power, the reduction in the target rotational speed of the rotary motor is calculated, including:

[0062] When the second predicted power is greater than the battery's maximum discharge power At that time, the current electric rotary power will be reduced accordingly, and the target speed of the rotary motor will decrease by [amount missing]. ;

[0063] When the second predicted power is less than the battery's maximum charging power At that time, the current electric rotary power will be reduced accordingly, and the target speed of the rotary motor will decrease by [amount missing]. ;

[0064] When the second predicted power is greater than the battery's maximum charging power and less than the battery's maximum discharging power. At that time, the target speed of the rotary motor decreases. 0;

[0065] Then, the required reduction in rotational speed for the current electric rotary valve is calculated.

[0066] Where T is the current torque of the rotary motor.

[0067] Furthermore, based on the final power requirement of the ISG and the target speed reduction of the electric rotary mechanism, corresponding control signals are converted into those for the ISG motor and electric rotary mechanism, including:

[0068] The target speed of the electric rotary mechanism is calculated by reducing the target speed of the electric rotary mechanism, and a control signal is generated based on the target speed of the electric rotary mechanism to drive the electric rotary mechanism to output the corresponding speed.

[0069] Based on the final power requirement of the ISG, the control torque of the ISG is calculated by dividing it by the current speed of the ISG motor. A control signal is then generated based on the control torque of the ISG to drive the ISG motor to output the corresponding control torque.

[0070] Furthermore, the target speed of the electric rotary mechanism is calculated based on the reduction in the target speed of the electric rotary mechanism, including:

[0071] According to the target speed reduction of the electric rotary mechanism Calculate the target speed of the electric rotary mechanism .

[0072] .

[0073] In the formula This is a request for the initial speed of the rotary motor calculated based on the handle opening.

[0074] Secondly, the present invention provides a power control system suitable for hybrid excavators, comprising:

[0075] The data sensing layer is used to collect and preprocess time-series data of handle opening, rotary motor speed, rotary motor torque, rotary motor bus voltage, and rotary motor bus current.

[0076] The LSTM network prediction layer, connected to the data perception layer, is used to input the preprocessed time series data into the trained LSTM neural network model to predict the power prediction value of the rotary motor at future times.

[0077] The power control layer, connected to the LSTM network prediction layer, is used to predict the battery power based on the predicted power value of the rotary motor, calculate the final power requirement of the ISG based on the battery power, and calculate the target speed reduction of the electric rotary mechanism based on whether the predicted battery power exceeds the power limit.

[0078] The execution layer, connected to the power control layer, is used to convert the final power demand of the ISG and the target speed reduction of the electric rotary mechanism into corresponding control signals for the ISG motor and electric rotary mechanism.

[0079] Furthermore, the data perception layer includes:

[0080] The sensor unit is used to collect the raw signals of the handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current.

[0081] The preprocessing unit is used to perform timestamp alignment, filtering, and normalization on the original signal.

[0082] Furthermore, the LSTM network prediction layer includes a trained LSTM neural network model, which takes multivariate time series data within a past time window as input and outputs a predicted value of the rotary motor power at a specified future time.

[0083] Furthermore, the LSTM neural network model includes an input layer, a first LSTM layer, a second LSTM layer, and an output layer connected in sequence;

[0084] The input layer is used to receive input data of shape (Ts, F), where Ts is the time step and F is the feature dimension;

[0085] The first LSTM layer is used to extract the temporal features of the input data, and its memory unit count is 64.

[0086] The second LSTM layer is used to further analyze the temporal features and output the feature vector of the final time step, and its memory unit number is 32;

[0087] The output layer is a fully connected layer used to map the feature vector into predicted values ​​of rotary motor power.

[0088] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0089] 1. Proactive control: By using an LSTM network to accurately predict the power of the rotary motor, the power demand of the battery can be anticipated in advance, realizing the transformation from "passive response" to "active adjustment".

[0090] 2. Dual adjustment mechanism: By calculating the first power increment, the battery power is stabilized within the ideal range. At the same time, by calculating the second power increment, the battery SOC is adjusted to near the target value. This dual protection ensures the battery's working state and effectively smooths out power and SOC fluctuations.

[0091] 3. Safety protection: By calculating the reduction in electric rotation speed, the electric rotation mechanism is limited in power control when the battery power is about to exceed the limit safety range, providing a last safety barrier for the system.

[0092] 4. Extend battery life: Through the above comprehensive control strategies, the current stress and cycle number of battery charging and discharging are significantly reduced, thereby effectively extending the battery's lifespan and reducing the user's total cost of ownership (TCO). Attached Figure Description

[0093] Figure 1 This is a schematic diagram of the power architecture of the hybrid excavator in an embodiment of the present invention.

[0094] Figure 2 This is a schematic diagram of the hybrid excavator power control system provided in an embodiment of the present invention.

[0095] Figure 3 This is a schematic diagram of the architecture of the LSTM neural network model in an embodiment of the present invention.

[0096] Figure 4 This is a schematic diagram of the internal structure of the LSTM memory cell in an embodiment of the present invention. Detailed Implementation

[0097] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0098] Example 1

[0099] This embodiment describes in detail a power control method suitable for hybrid excavators. The process of this method is as follows: Figure 2 As shown, applied to Figure 1 The hybrid excavator power architecture shown includes the following steps:

[0100] S101: Collect and preprocess time series data.

[0101] The system collects raw signals of handle opening (%), rotary motor speed (rpm), rotary motor torque (N•m), rotary motor bus voltage (V), and rotary motor bus current (A) using sensors installed on the vehicle.

[0102] Preprocessing of the raw signal includes:

[0103] a) Timestamp alignment: Ensure that all sensor data is synchronized over time.

[0104] b) Filtering: A first-order low-pass filter is used to filter each data sequence to remove high-frequency noise.

[0105] c) Normalization: The Min-Max normalization algorithm is used to map each feature value to the interval [0,1].

[0106] S102: Predicts the power of the rotary motor.

[0107] The preprocessed time series data is input into a pre-trained LSTM neural network model to predict the power of the rotary motor at a future time (e.g., 1 second later).

[0108] The structure of the LSTM neural network model is as follows: Figure 3 As shown, the model includes an input layer, a first LSTM layer (64 units), a second LSTM layer (32 units), and an output layer (linear activation function). The model is trained using historical data as samples and mean squared error (MSE) as the loss function, learning the mapping relationship from the input sequence to future power values. The internal computation flow of the LSTM memory unit is as follows... Figure 4 As shown.

[0109] S103: Estimate the first predicted power of the battery.

[0110] Future rotary motor power predicted by LSTM Based on the principle of equal power on the bus, the first predicted power of the battery can be calculated as follows:

[0111] ;

[0112] In the formula The current ISG motor power can be calculated by multiplying the motor bus voltage and current.

[0113] S104: Calculate the power increment of the ISG motor.

[0114] This step includes the following sub-steps:

[0115] S1041: Calculate power limits.

[0116] 1. Calculate the ideal charging power limit of the battery based on the battery's SOC and temperature. Ideal discharge power limit Maximum charging power limit Maximum discharge power limit The charging power is negative, and the discharging power is positive.

[0117] The calculation of the above values ​​should refer to the battery's charge and discharge power MAP. For example, for a certain battery with a SOC of 60% and a temperature of 25℃, the calculated ideal charging power limit is -50kW, the ideal discharging power is 50kW, the extreme charging power is -100kW, and the extreme discharging power is 100kW.

[0118] 2. Calculate the maximum available power of the ISG motor based on the power limits of the hydraulic system and the engine. Minimum available power The maximum available power is a positive value, indicating that the power is driving; the minimum available power is a negative value, indicating that the power is generating.

[0119] In the series connection of the engine, ISG motor, and hydraulic system, the power balance relationship is as follows:

[0120] ;

[0121] This represents the current power of the hydraulic system, which can be calculated by measuring the pressure and flow rate of the hydraulic pump; it is a positive power.

[0122] This represents the current engine power, and its power range is... This is related to the external characteristics of the engine.

[0123] Therefore, the minimum usable power of the ISG motor during normal operation is calculated as follows: The maximum available power is ;

[0124] This is the minimum power of the ISG motor (negative value when generating electricity). The maximum power of the ISG motor (positive when driven) and these two values ​​depend on the external characteristics of the motor.

[0125] S105: Calculate the additional power increment of the ISG motor.

[0126] Based on the battery's first predicted power The difference between the power output and the ideal charging / discharging power is used to adjust the first target power increment of the ISG motor. By adjusting the power of the ISG motor, the future power of the battery can be limited as much as possible to the ideal charging and discharging power range.

[0127] when When the ideal charging power limit is between the ideal discharging power limit and the ideal charging power limit, i.e. The first ISG motor target power increment The ISG motor maintains its current power output.

[0128] when When the ideal discharge power limit is exceeded, i.e. The first ISG motor target power increment , When the value is negative, the ISG motor reduces its current power output.

[0129] when When the charging power is below the ideal charging power limit, i.e. The first ISG motor target power increment , When the value is positive, the ISG motor increases its current power output.

[0130] The second target power increment of the ISG motor is adjusted based on the difference between the current SOC and the target SOC of the battery. By adjusting the motor power, the battery SOC can be controlled to be near the target SOC value.

[0131] The basis for calculating the SOC increment is that when the SOC is higher than the target SOC, at this point... When the value is positive, the ISG motor increases its current power output, either generating less electricity or consuming more; when the SOC is lower than the target SOC, at this time... When the value is negative, the ISG motor reduces its current power output, generating more electricity or consuming less electricity;

[0132] S106: Calculate and limit the final power demand of the ISG.

[0133] Based on the first target power increment of the ISG motor Second target power increment Calculate the final power demand of the ISG And based on the minimum available power of the ISG motor. Maximum available power Limit this value.

[0134] Power demand ;

[0135] And limit the size of this value to [ ].

[0136] S107: Determine and calculate the speed reduction.

[0137] This step is a safety protection step, including:

[0138] The second predicted power of the battery is updated based on the final power demand of the ISG. .

[0139]

[0140] When the rotary motor is driving and braking, there will be peak discharge and charging power, which may cause the second predicted power of the battery to exceed the battery's limit charge and discharge power. At this time, the speed of the rotary motor should be limited accordingly to reduce the electric rotary power.

[0141] When it exceeds the battery's maximum discharge power or less than the battery's maximum charging power At this time, the current electric rotation power should be reduced accordingly, by a reduction value of [value missing]. , or Then, the required reduction in rotational speed for the current electric rotary valve is calculated. .

[0142] According to the target speed reduction of the electric rotary mechanism Calculate the target speed of the electric rotary mechanism .

[0143] .

[0144] In the formula The initial speed request of the rotary motor is calculated based on the handle opening degree and is usually linearly proportional to the opening degree.

[0145] S108: Generate and send control signals.

[0146] Based on the final power requirement of the ISG motor obtained in step S106 and the target speed reduction obtained in step S107, corresponding control signals (such as CAN messages) are generated and sent to the ISG motor controller and the electric rotary motor controller for execution, thereby achieving smooth control of battery power and SOC.

[0147] Example 2:

[0148] This embodiment provides a power control system suitable for hybrid excavators, which is used to execute the method described in Embodiment 1.

[0149] This invention discloses a power control system for a hybrid excavator, applicable to the power architecture of a hybrid excavator, such as... Figure 1 As shown. Based on this power architecture, the designed power control system mainly includes the following:

[0150] 1. Data Awareness Layer

[0151] Sensors installed on the vehicle collect data on handle opening (%), rotary motor speed (rpm), rotary motor torque (N•M), rotary motor bus voltage (V), and rotary motor bus current (A), and then filter and normalize the data.

[0152] 2. LSTM network prediction layer

[0153] The data processed by the data sensing layer is fed into the trained LSTM network. The LSTM network performs electrogyratory power prediction based on the time series data and outputs the predicted power value.

[0154] 3. Power Control Layer

[0155] By predicting the electric slewing power, and then predicting the battery power, in order to control the battery charging and discharging power and SOC within the ideal range, the power increment of the ISG motor is calculated, and the target speed reduction of the electric slewing mechanism is calculated based on whether the predicted battery power exceeds the allowable limit power.

[0156] 4. Execution layer

[0157] Based on the target power increment of the first and second ISG motors and the target speed reduction of the electric rotary mechanism output by the power control layer, the corresponding control signals for the ISG motors and electric rotary mechanisms are converted.

[0158] The data perception layer includes:

[0159] The sensor unit is used to collect the raw signals of the handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current.

[0160] The preprocessing unit is used to perform timestamp alignment, filtering, and normalization on the original signal.

[0161] The LSTM network prediction layer includes a trained LSTM neural network model. The input of the model is multivariate time series data within a past time window, and the output is the predicted value of the rotary motor power at a specified future time.

[0162] The LSTM neural network model includes an input layer, a first LSTM layer, a second LSTM layer, and an output layer connected in sequence.

[0163] The input layer is used to receive input data of shape (Ts, F), where Ts is the time step and F is the feature dimension;

[0164] The first LSTM layer is used to extract the temporal features of the input data, and its memory unit count is 64.

[0165] The second LSTM layer is used to further analyze the temporal features and output the feature vector of the final time step, and its memory unit number is 32;

[0166] The output layer is a fully connected layer used to map the feature vector into predicted values ​​of rotary motor power.

[0167] Example 3:

[0168] This embodiment provides a power control method suitable for hybrid excavators.

[0169] This invention discloses a power control system for a hybrid excavator, applicable to the power architecture of a hybrid excavator, such as... Figure 1 As shown. Based on this power architecture, the designed power control system mainly includes the following:

[0170] Data perception layer: Through sensors installed on the vehicle, the handle opening degree (in %), rotary motor speed (in rpm), rotary motor torque (in N•M), rotary motor bus voltage (in V), and rotary motor bus current (in A) are collected, and the data are filtered and normalized.

[0171] LSTM network prediction layer: The data processed by the data perception layer is provided to the trained LSTM network. The LSTM network performs electrogyratory power prediction based on the time series data and outputs the power prediction value.

[0172] Power control layer: By predicting the electric slewing power, the battery power is predicted. In order to control the battery charging and discharging power and SOC within the ideal range, the power increment of the ISG motor is calculated, and the target speed reduction of the electric slewing mechanism is calculated based on whether the predicted battery power exceeds the allowable limit power.

[0173] Execution layer: Based on the final power demand of the ISG output by the power control layer and the reduction of the target speed of the electric rotary mechanism, it converts the power demand of the ISG into corresponding control signals for the ISG motor and electric rotary mechanism.

[0174] Based on the above power control system, the present invention designs a power control method suitable for hybrid excavators, which mainly includes the following steps:

[0175] S1: Sensor data acquisition and preprocessing.

[0176] 1. Data Collection

[0177] Using sensors installed on the vehicle, five physical features were collected: handle opening (%), rotary motor speed (rpm), rotary motor torque (N•M), rotary motor bus voltage (V), and rotary motor bus current (A). The time series of each feature was collected as a dataset.

[0178] Calculate the electric rotation power for each row of data in the dataset, and calculate the electric power of the electric rotation mechanism at the current moment based on the motor bus voltage and current.

[0179] 2. Pretreatment

[0180] a) Align timestamps to ensure that the data collected by the sensors is consistent over time.

[0181] b) Eigenvalue filtering: First-order low-pass filtering is performed on the time series data of each sensor to remove high-frequency noise.

[0182] c) Eigenvalue normalization: In order to eliminate the influence of different units and dimensions of eigenvalues, the collected sensor data is normalized by Min-Max, and the range of each eigenvalue is limited to [0 1].

[0183] S2: Construction of LSTM network training samples

[0184] Predict the power value for a future specified time period based on sensor data from the past.

[0185] 1. Input sample definition:

[0186] The time window width is =60, corresponding to a 6-second time window, sampling frequency. It is 10Hz.

[0187] Feature Dimension =5, corresponding to the 5 physical quantities collected above.

[0188] Then the dimension of each input sample is ( , ).

[0189] The time window sliding step is =10, assuming the total amount of data collected is If the time is seconds, then the total number of samples N is

[0190] ;

[0191] 2. Output sample definition:

[0192] To predict future power values, the electric slewing power of the dataset for the next P steps is used as the output sample, based on the timestamp of the sample. Due to the slewing characteristics of excavators, the electric slewing power changes instantaneously, so the prediction time should not be too long; 1 second is recommended.

[0193] S3: LSTM Network Design and Training

[0194] The LSTM network structure is shown in the figure.

[0195] 1. Input Layer

[0196] The function of this input layer is to receive the sequence data of the above five physical quantities collected by the original sensor, without performing any calculations. The shape of this input layer is ( Input at each moment The dimension is F.

[0197] 2. First LSTM layer (LSTM Layer 1)

[0198] The main function of this LSTM layer is to extract the short-term and long-term dependencies between various physical quantities and faults, and to extract temporal features. The LSTM is designed with 64 memory cells.

[0199] The internal structure of each LSTM memory cell is as follows: Figure 4 As shown, for each time step, the hidden state and memory unit are calculated step-by-step, and the internal calculation formula is as follows:

[0200] Forgotten Gate:

[0201] Input Gate:

[0202] Candidate memory units:

[0203] Update memory units:

[0204] Output gate:

[0205] Update hidden status:

[0206] In the formula It is the sigmoid activation function;

[0207] The tanh activation function;

[0208] For trainable parameter matrices and biases;

[0209] - Element-by-element multiplication;

[0210] This layer retains the output of each time step for use by the lower layer; the final output at each time step is... Its shape is ( ).

[0211] 3. LSTM Layer 2

[0212] The LSTM layer in this stage receives the output from the first layer and further analyzes and makes decisions based on the timing characteristics. The LSTM layer is designed with 32 memory cells.

[0213] The formula for calculating the internal memory cells of this LSTM layer is the same as that of the first LSTM layer.

[0214] This layer only retains the output of the last time step for use by the lower layers; the final output at each time step... Its shape is (1, 32).

[0215] 4. Output Layer

[0216] This layer maps the features extracted by the LSTM layer to the rotary motor power space. It has a single unit, representing the predicted rotary motor power value at the next 1 second. The calculation formula for this layer is:

[0217] ;

[0218] In the formula It is a linear activation function, suitable for continuous values ​​such as power values.

[0219] This represents the trainable parameter matrix and biases for this layer.

[0220] 5. Loss Function

[0221] During the sample training phase, the training objective is to determine the optimization direction of the network model through the loss function. Using the mean squared error function as the loss function is beneficial for handling large errors and capturing peak power, as defined below:

[0222] ;

[0223] In the formula, The number of samples; Output for training purposes; The output of the network prediction;

[0224] During training, iterative optimization using samples, such as selecting Adam to optimize the parameter matrices and biases of each network layer, is achieved. Approaching 0.

[0225] S5: Battery First Predicted Power Calculation

[0226] Future rotary motor power predicted by LSTM Based on the principle of equal power on the bus, the first predicted power of the battery can be calculated as follows:

[0227] ;

[0228] In the formula The current ISG motor power can be calculated by multiplying the motor bus voltage and current.

[0229] S6: Power Limit Calculation

[0230] Calculate the ideal charging power limit of the battery based on the battery's state of charge (SOC) and temperature. Ideal discharge power limit Maximum charging power limit Maximum discharge power limit The charging power is negative, and the discharging power is positive.

[0231] The calculation of the above values ​​should refer to the battery's charge and discharge power MAP. For example, for a certain battery with a SOC of 60% and a temperature of 25℃, the calculated ideal charging power limit is -50kW, the ideal discharging power is 50kW, the extreme charging power is -100kW, and the extreme discharging power is 100kW.

[0232] The maximum available power of the ISG motor is calculated based on the hydraulic system power and engine power limits. Minimum available power The maximum available power is a positive value, indicating that the power is driving; the minimum available power is a negative value, indicating that the power is generating.

[0233] In the series connection of the engine, ISG motor, and hydraulic system, the power balance relationship is as follows:

[0234] ;

[0235] This represents the current power of the hydraulic system, which can be calculated by measuring the pressure and flow rate of the hydraulic pump; it is a positive power.

[0236] This represents the current engine power, and its power range is... This is related to the external characteristics of the engine.

[0237] Therefore, the minimum usable power of the ISG motor during normal operation is calculated as follows: The maximum available power is ;

[0238] This is the minimum power of the ISG motor (negative value when generating electricity). The maximum power of the ISG motor (positive when driven) and these two values ​​depend on the external characteristics of the motor.

[0239] S7: ISG First Target Power Increment Calculation

[0240] Based on the battery's first predicted power The difference between the power output and the ideal charging / discharging power is used to adjust the first target power increment of the ISG motor. By adjusting the power of the ISG motor, the future power of the battery can be limited as much as possible to the ideal charging and discharging power range.

[0241] when When the ideal charging power limit is between the ideal discharging power limit and the ideal charging power limit, i.e. The first ISG motor target power increment The ISG motor maintains its current power output.

[0242] when When the ideal discharge power limit is exceeded, i.e. The first ISG motor target power increment , When the value is negative, the ISG motor reduces its current power output.

[0243] when When the charging power is below the ideal charging power limit, i.e. The first ISG motor target power increment , When the value is positive, the ISG motor increases its current power output.

[0244] S8: ISG Second Target Power Increment Calculation

[0245] The second target power increment of the ISG motor is adjusted based on the difference between the current SOC and the target SOC of the battery. By adjusting the motor power, the battery SOC can be controlled to be near the target SOC value.

[0246] The basis for calculating the SOC increment is that when the SOC is higher than the target SOC, at this point... When the value is positive, the ISG motor increases its current power output, either generating less electricity or consuming more; when the SOC is lower than the target SOC, at this time... When the value is negative, the ISG motor reduces its current power output, generating more electricity or consuming less electricity;

[0247] The target SoC can be 60%. The value can be obtained by looking up Table 1 as shown below, with the unit being Kw:

[0248] Table 1 Mapping Table of Second ISG Target Power Increment and Second Power Increment

[0249]

[0250] S9: ISG Final Power Demand Calculation

[0251] Based on the first target power increment of the ISG motor Second target power increment Calculate the final power demand of the ISG And based on the minimum available power of the ISG motor. Maximum available power Limit this value.

[0252] Power demand ;

[0253] And limit the size of this value to [ ].

[0254] S10: Battery Second Predicted Power Calculation

[0255] The second predicted power of the battery is updated based on the final power demand of the ISG. .

[0256]

[0257] S11: Calculation of target speed reduction of rotary motor

[0258] When the rotary motor is driving and braking, there will be peak discharge and charging power, which may cause the second predicted power of the battery to exceed the battery's limit charge and discharge power. At this time, the speed of the rotary motor should be limited accordingly to reduce the electric rotary power.

[0259] When it exceeds the battery's maximum discharge power or less than the battery's maximum charging power At this time, the current electric rotation power should be reduced accordingly, by a reduction value of [value missing]. , or Then, the required reduction in rotational speed for the current electric rotary valve is calculated. .

[0260] S12: Calculation of the final required speed of the rotary motor

[0261] According to the target speed reduction of the electric rotary mechanism Calculate the target speed of the electric rotary mechanism .

[0262] .

[0263] In the formula The initial speed request of the rotary motor is calculated based on the handle opening degree and is usually linearly proportional to the opening degree.

[0264] By using an LSTM network to predict the power of the rotary motor, the future power of the power battery can be predicted. By adjusting the power of the ISG motor, the power battery power and SOC can be controlled within the ideal range, reducing battery charging and discharging, which helps to extend battery life.

[0265] The parameters in the LSTM network designed in this article are only one specific embodiment. Changing other layers, such as the number of units or the number of layers, is also within the scope of this technology.

[0266] The power regulation method for ISG described in this article is just one specific embodiment. Changing the data or using other formulas also falls within the scope of this technology.

[0267] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0268] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0269] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0270] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0271] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A power control method suitable for hybrid excavators, characterized in that, The method includes: Collect and preprocess time series data; The preprocessed time series data is input into a trained LSTM neural network model to predict the power of the rotary motor at a future time. The input of the LSTM neural network model is multivariate time series data within a past time window, and the output is the predicted power of the rotary motor at a specified future time. The battery power is predicted based on the predicted power value of the rotary motor, the final required power of the ISG is calculated based on the battery power, and the target speed reduction of the electric rotary mechanism is calculated based on whether the predicted battery power exceeds the power limit. Based on the final power requirement of the ISG and the target speed reduction of the electric rotary mechanism, the corresponding control signals for the ISG motor and electric rotary mechanism are converted. The battery power is predicted based on the predicted power value of the rotary motor. The final power requirement of the ISG is calculated based on the battery power. The target speed reduction of the electric rotary mechanism is calculated based on whether the predicted battery power exceeds the power limit, including: Calculate the first predicted power of the battery based on the predicted power value of the rotary motor; Obtain the power limit of the ISG motor; Based on the first predicted power of the battery, calculate the first target power increment of the ISG and the second target power increment of the ISG. Calculate the final power requirement of the ISG based on the power limit of the ISG motor, the first target power increment of the ISG, and the second target power increment of the ISG. Update the second predicted power of the battery based on the final power demand of the ISG; The reduction in the target speed of the rotary motor is calculated based on the second predicted power of the battery.

2. The power control method for hybrid excavators according to claim 1, characterized in that, The time series data includes time series data of handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current; The collection and preprocessing of time series data includes: The raw signals of handle opening degree, rotary motor speed, rotary motor torque, rotary motor bus voltage and rotary motor bus current are collected by sensors. The original signal is then subjected to timestamp alignment, filtering, and normalization.

3. The power control method for hybrid excavators according to claim 1, characterized in that, The training method for the LSTM neural network model includes: Construct a training dataset, wherein the input samples of the dataset are the time series data within a past time window, and the output samples are the actual rotary motor power values ​​at a specified future time. Using mean squared error as the loss function, Adam is selected to optimize the parameter matrix and bias of each layer of the network, and the constructed LSTM neural network model is trained.

4. The power control method for hybrid excavators according to claim 1, characterized in that, The first predicted power of the battery is calculated based on the predicted power value of the rotary motor, including: ； in, The first predicted power of the battery, For the predicted rotary motor power, This represents the current power of the ISG motor.

5. The power control method for hybrid excavators according to claim 4, characterized in that, Obtain the power limits for the ISG motor, including: Based on the battery's SOC and temperature, and referring to the battery's charge / discharge power MAP, the ideal charging power limit of the battery is calculated. Ideal discharge power limit Maximum charging power limit Maximum discharge power limit The charging power is negative, and the discharging power is positive. The maximum available power of the ISG motor is calculated based on the hydraulic system power and engine power limits. Minimum available power The maximum available power is a positive value, indicating that the power is driving; the minimum available power is a negative value, indicating that the power is generating. Calculate the minimum available power of the ISG motor during normal operation. The maximum available power is ; in, This is the minimum power of the ISG motor. These two values ​​represent the maximum power of the ISG motor and depend on the motor's external characteristics. The current power of the hydraulic system is calculated by measuring the pressure and flow rate of the hydraulic pump, and is a positive power. This represents the current engine power, and its power range is... .

6. The power control method for hybrid excavators according to claim 5, characterized in that, Calculating the first target power increment of the ISG includes: According to the first predicted power of the battery With respect to the ideal charging power limit of the battery and ideal discharge power limit The relationship is used to calculate the first power increment. The relationship is as follows: when When the ideal charging power limit is between the ideal discharging power limit and the ideal charging power limit, i.e. The first ISG motor target power increment The ISG motor maintains its current power output; when When the ideal discharge power limit is exceeded, i.e. The first ISG motor target power increment , When the value is negative, the ISG motor reduces its current power output; when When the charging power is below the ideal charging power limit, i.e. The first ISG motor target power increment , When the value is positive, the ISG motor increases its current power output; Calculating the second target power increment of the ISG includes: Based on the difference between the battery's current SOC and the target SOC, a preset mapping table is consulted to obtain the second power increment. .

7. The power control method for hybrid excavators according to claim 6, characterized in that, Based on the power limit of the ISG motor, the first target power increment of the ISG, and the second target power increment of the ISG, calculate the final power requirement of the ISG, including: Based on the first target power increment of the ISG motor Second target power increment Calculate the final power demand of the ISG : ； Based on the minimum available power of the ISG motor Maximum available power Limit this value: reduce the final power demand. The value is limited to [ , ].

8. The power control method for hybrid excavators according to claim 1, characterized in that, Based on the ISG's final power demand, the second predicted power of the battery is updated, including: The second predicted power of the battery is updated based on the final power demand of the ISG. : ； in, For the final required power, The predicted power of the rotary motor; Based on the battery's second predicted power, the reduction in the target speed of the rotary motor is calculated, including: When the second predicted power is greater than the battery's maximum discharge power At that time, the current electric rotary power will be reduced accordingly, and the target speed of the rotary motor will decrease by [amount missing]. ; When the second predicted power is less than the battery's maximum charging power At that time, the current electric rotary power will be reduced accordingly, and the target speed of the rotary motor will decrease by [amount missing]. ; When the second predicted power is greater than the battery's maximum charging power and less than the battery's maximum discharging power. At that time, the target speed of the rotary motor decreases. 0; Calculate the required reduction in rotational speed for the current electric rotary valve. ; Where T is the current torque of the rotary motor.

9. A control system applying the power control method for hybrid excavators as described in claim 1, characterized in that, include: The data sensing layer is used to collect and preprocess time-series data of handle opening, rotary motor speed, rotary motor torque, rotary motor bus voltage, and rotary motor bus current. The LSTM network prediction layer, connected to the data perception layer, is used to input the preprocessed time series data into the trained LSTM neural network model to predict the power prediction value of the rotary motor at future times. The power control layer, connected to the LSTM network prediction layer, is used to predict the battery power based on the predicted power value of the rotary motor, calculate the final power requirement of the ISG based on the battery power, and calculate the target speed reduction of the electric rotary mechanism based on whether the predicted battery power exceeds the power limit. The execution layer, connected to the power control layer, is used to convert the final power demand of the ISG and the target speed reduction of the electric rotary mechanism into corresponding control signals for the ISG motor and electric rotary mechanism.

Images