150 results about "Hyperbolic tangent function" patented technology

The present invention relates to a non-affine uncertain system self-adaptive control method with a range restraint. In combination with the characteristic that system input and states are subject to the range restraint, based on a mean value theorem, a non-affine system is converted into a time varying system (a strict feedback system) with a linear structure, and a varying interval of time-varying uncertain parameters of the time varying system is obtained. On the basis, a parameter adaptive projection technique is utilized to carry out online estimation on bounded uncertain time-varying parameters and external disturbances, and a parameter estimation error is compensated by using the nonlinear dynamic damping technology. Based on the nonlinear mapping technology, a restraint amount is mapped to the whole real number space to solve a state restraint problem of the system, thus to ensure the state is always within a restraint range. A hyperbolic tangent function and the Nussbaum gain technology are simultaneously utilized to solve the problem that system input is subject to the range restraint or the problem of input saturation and the potential problem of singular values of a controller, and the controller is designed in combination with a method of inversion, thereby solving the problem of self-adaptive control of the non-affine uncertain system with the range restraint.

The invention provides a self-disturbance control method for a permanent magnet synchronous motor, and belongs to the technical field of permanent magnet synchronous motor control. The method comprises the steps of: designing a hyperbolic tangent function expansion state observer, adopting the saturation characteristics itself of the hyperbolic tangent function to effectively suppress a differential peak appearing in a traditional expansion state observer so as to improve the system control performance. The non-linear feedback control rate based on the new adaptive sliding mode control is proposed which is characterized by designing a new adaptive shifting index approach law which introduces a first-order norm of the state variable, and an approaching speed is adaptively regulated according to the distance of the state variable from the balance point to reduce the dependence of the self-disturbance control on the parameter robustness so as to achieve the high-precision and high robustness control of a permanent magnet synchronous motor. When the system is disturbed and a load fluctuates, the rotation speed is rapidly tracked, the response speed of the system is improved, the systemovershoot and the steady state static difference are reduced, and the system robustness is greatly enhanced.

The invention provides a sub pixel edge based circle center detecting method. The method includes steps of 1, utilizing a Canny operator to performing rough positioning on an image edge, acquiring the gradient magnitude and gradient direction of integer pixel point edge; 2, establishing a positive direction taking the integer pixel point as the origin and the taking the gradient direction as the x axis, taking 7 pixel points in central symmetry in the gradient direction, wherein the symmetry center which is the integer pixel point is among the 7 pixel points, performing gray level fitting on the 7 pixel points by adopting a hyperbolic tangent function in a formula of I(x)=-b / 2 tanh(p(x-R))+h+b / 2, wherein I(x) indicates the gray level in x position, x indicates the distance between the pixel point to the origin, h indicates background gray level, b indicates the gray level difference value of the edge and the background, p is a fuzzy factor, and R is the edge position; and 3, obtaining a sub pixel edge point according to the formula in the step 2, performing least square fit on the sub pixel edge point according to a circle and obtaining circle center coordinates.

An inverted pendulum self-adaptive iterative learning inversion control method is characterized in that: aiming at an inverted pendulum system containing unknown input saturation, a self-adaptive iterative learning inversion controller is designed through utilization of a neural network and an inversion control method in combination with self-adaptive iterative learning control; the construction of an integral lyapunov function solves the control problem caused by derivation of a unknown gain function; based on the median theorem, a hyperbolic tangent function is adopted to approximate an input saturation term; then, a radial basis function neural network is adopted to approximate and compensate uncertain unknown items of a system, and two combined self-adaptive laws are adopted to updatethe weight of the neural network and the bound of estimation errors. Under the condition that the system has input saturation, the invention provides the control method which can compensate the unknown uncertainty of the system, solve the control problem caused by derivation of the unknown gain function and realize two-norm convergence of a tracking error of the system to be near zero within limited iteration times.

The present invention discloses a method for reducing PAPR of OFDM system. The method replaces traditional Mu law companding function by hyperbolic tangent function, to ensure the value obtained after companding the data of the original signal be within a certain range, the concrete process is that as follows: in transmission terminal, to calculate the time field OFDM signal after IFFT so as to obtain average amplitude, and confirm companding parameter k and amplitude modulation parameter Alpha based on the demand of OFDM system; according to the average amplitude and Alpha as well as k, to implement companding transformation to the OFDM signal with hyperbolic tangent function mode, then to transmit to the channel after parallel-serial conversion and filtering; in the receiving terminal, to calculate the received OFDM signal so as to obtain the average amplitude of the received signal, and to limit the amplitude of the OFDM signal based on threshold; according to average amplitude and alpha as well as k, to implement companding inverse transformation to the OFDM signal with hyperbolic tangent companding inverse transformation function mode, so as to resume the original signal. The present invention has the advantage of reducing PAPR well, and can be used to treat the PAPR in the fourth generation OFDM radio communication system.

The invention discloses a bridge type crane system APD-SMC controller, a bridge type crane system and a control method. The bridge type crane system comprises a trolley, a bridge frame and a load; a target position of the trolley is defined as pd, state vectors are formed by displacement x of the trolley and a swing angle theta of the load relative to a vertical direction, and therefore a state equation of the bridge type crane system can be obtained; a composite function xi x, an auxiliary function, a sliding mode surface s and a hyperbolic tangent function are introduced; a bridge type crane system APD-SMC controller is therefore obtained. According to the bridge type crane system APD-SMC controller, the bridge type crane system and the control method, a coupling relation between the displacement of the trolley and the swing angle of the load is strengthened, and therefore transient control performance of the system is improved.

The invention belongs to the field of control over a permanent magnet synchronous motor, and relates to a speed regulating control strategy for a sensorless permanent magnet synchronous motor. The speed regulating control strategy is characterized in that a three-phase current of the permanent magnet synchronous motor is detected, the current under an alpha and beta coordinate system is obtained through Clarke conversion, and the current under the dq coordinate system is obtained through Park conversion; a sliding-mode observer is designed according to a mathematic model of the permanent magnet synchronous motor under the alpha and beta two-phase static coordinate system; a hyperbolic tangent function is used as a switching function; a sliding-mode face is selected; the estimation value of the rotor position is obtained through counter electromotive force; a speed estimator is adopted for estimating the speed; a Lyapunov function is selected, and an adaptive law is obtained through the stability condition; According to the speed regulating control strategy, the structural switching discontinuity is relieved, the buffeting is obviously weakened, the rapidness, the robustness and the dynamic performance are improved, when an error exists in the estimation value of the rotor position, the speed estimation error is made to be within the controllable range, and the rotating speed estimation precision is improved.

The invention relates to a hypersonic aircraft robust tracking control method based on a terminal sliding mode, and solves the problems of complex control model, poor robustness and lack of consideration of controller input limitation of an existing aircraft. A hypersonic aircraft input / output linearization model is provided, and is converted into a second-order system model by introducing an error auxiliary variable; for the conditions that unknown upper bound exists in system interference and an actuator has no input saturation, an adaptive fast terminal sliding mode controller is designed based on a fast non-singular terminal sliding mode surface, thereby ensuring that the sliding mode surface is actually convergent in finite time; and by introducing a hyperbolic tangent function and astructural auxiliary system, an anti-saturation adaptive fast terminal sliding mode controller is designed, thereby meeting requirements of physical constraints of a hypersonic aircraft and meanwhile,ensuring that the system sliding mode surface is convergent in finite time. The method is applied to the aircraft field.

The invention discloses a reversed-U-type dielectric waveguide wideband directional coupler used for a millimeter wave circuit, which is composed of two medium waveguide transmission lines and an earthing metal plate, wherein two medium waveguide transmission lines close to each other are oppositely bent, arranged separately and symmetrically at two sides of an axle wire and fixed on the earthing metal plate, so as to form an X-type structure; the coupler is characterized in that the reversed-U-type medium waveguide is used for reaching of the wideband plat coupling of the coupling section and the transition section of a coupler, and the waveguide axes of the transition section adopts hyperbolic tangent line type; the fluctuation of a coupling coefficient doesn't exceed plus or minus 0.5Db inside the frequency coverage of 6GHz, and the flat coupling property can be achieved inside wider bandwidth. Compared with the traditionally directional coupler which is completely composed of a medium mirror image line, the inverted-U-type dielectric waveguide wideband directional coupler has the advantages that transition waveguide section of the hyperbolic tangent line type is used, the metal side wall of the prior coupler which is vertical to ground in the transition waveguide section is unnecessarily used, radiation loss and metal loss are reduced, the volume of the coupler is decreased and manufacture complexity and cost are reduced, and can be conveniently produced in large quantity.

The invention discloses an anti-saturation boundary control method for a marine flexible riser system. Based on the Lyapunov theory and a back-stepping method, a smooth hyperbolic tangent function, anauxiliary system and a Nussbaum function are introduced to establish a boundary control rate of the marine flexible riser system. The control parameters in the boundary control rate are obtained through pre-training, and then the partial derivatives of the three parameters at the moment t are obtained through calculation of historical data according to the actually-measured shear force, inclination and displacement of the top of a riser, finally, the boundary control rate at the moment t is calculated based on the obtained actual parameters, and a driving device applies an acting force to thetop of the riser according to the boundary control rate. The invention further discloses an anti-saturation boundary control device for a marine flexible riser system, a storage medium and a server.The method and the device can solve the problem of tremor caused by using a sign function to limit the control input, can effectively control the vibration of the flexible riser, and are few in measurement parameters and easy to implement.

The invention discloses a bridge crane energy coupling controller and control method with initial input constraint. A generalized load horizontal displacement signal is established, so that the coupling relation among trolley movement, lifting hook swinging and swinging around a lifting hook of a load is enhanced, and the transient state control performance of the controller is improved. A hyperbolic tangent function is introduced into a control rate, and initial drive force is lowered greatly by means of the control method, so that smooth starting of a trolley is guaranteed, and great swinging of the load is avoided. Strict mathematical analysis is carried out on the stability of a closed-loop system at a balance point by means of a Lyapunov method and a LaSalle invariance principle. It is indicated by the results that the control algorithm has good control performance and has very high robustness on different load mass, rope lengths, target positions and external disturbances.

The present invention relates to an illumination-robust facial image local texture enhancement method includes the following steps that: logarithmic transformation is performed on the gray value of an inputted original face image I, so that a logarithmic transformation result image I' can be obtained; Gaussian differential filtering and bilateral differential filtering are performed on the logarithmic transformation result image I', so that differential filtering result images IDoG and IDoB are obtained, image information fusion is performed on IDoG and IDoB, so that a fusion result image I" is obtained; and the fusion result image I" is divided into sub-image blocks, grayscale-equalization processing is performed on each sub-image block by means of a mean normalization method, and the sub-image blocks are spliced according to division positions, and then the pixel gray value range of a spliced image is compressed by means of a hyperbolic tangent function, and an image is outputted. With the method of the invention adopted, facial images imaged under different illumination conditions can be processed; illumination influence can be eliminated; face local texture information can be enhanced; and recognition accuracy in face recognition application can be improved. The method has the advantages of low algorithm complexity and high light robustness.

The invention relates to a rolling bearing residual life prediction method, which predicts the life of a rolling bearing by using vibration signals of the rolling bearing through training a multi-scale convolutional neural network model, and comprises the following steps: carrying out an accelerated degradation experiment by using a plurality of unused bearings to obtain full-life vibration signals of the bearings; converting the service life of the bearing into a health index by using an inverse hyperbolic tangent function; establishing a multi-scale convolutional neural network model, and training the model by using the obtained data; measuring a vibration signal of the rolling bearing with the life to be predicted by using an acceleration sensor; inputting the obtained vibration signalinto a trained multi-scale convolutional neural network model to obtain a health index of the rolling bearing with the life to be predicted; and converting the obtained health index into the residuallife of the rolling bearing with the life to be predicted. The invention aims to provide a rolling bearing residual life prediction method which can efficiently and accurately predict the residual life of a rolling bearing under complex actual working conditions.

The invention discloses a design method of aeronautical engine integral tangent fuzzy self - adaptive sliding mode controller, which comprises the following steps: obtaining a model near a steady point to output data according to a nonlinear component level model of an aeroengine system, establishing a linear model of the aeroengine by using a least squares method; defining a tracking error of the aeroengine system, a reference signal for a given system, and assuming that a derivative of a reference signal is 0; designing a hyperbolic tangent plane by using a hyperbolic tangent function similar to a function of a saturation function; saturating an error for a case where the error is large, and enlarging a system error for a case where the error is small; improving a dynamic quality of a system approach motion by using an exponential approaching law; utilizing an adaptive control method to realize an adaptive estimation of an uncertainty upper bound, and the uncertainty upper bound is obtained; designing a sliding mode control law by means of the Lyapunov function verifying a stability of the system.

This invention puts forward a simplified algorithm of two-layer weighted parallel interference cancellation algorithm which greatly reduces operation volume which keeping its performance. Since it relates to hyperbolic tangent operation in the two layer weighted parallel interference cancellation algorithm it nearly can't by realized. This invented simplified algorithm is to use sectional linear judgement method or looking-up table method to replace the hyperbolic tangent judgement. The substance is to use sectional linear judgement function L(x) or judgement function T(x) of looking-up tableto approach the hyperbolic tangent function tanb(x).

The invention discloses a partial saturation adaptive controller of a bridge crane, a control system and a control method. The control method comprises the following steps: wx online estimation w(^)x and wl online estimation w(^)l are input to a neural network for dynamic learning; w(^)x and w(^)l' undergoes dynamic output by the neural network and are transmitted to the partial saturation adaptive controller; and finally the partial saturation adaptive controller outputs driving force which can guarantee smooth start-up of the bridge crane system, wherein wx and wl are both auxiliary vectors. By introducing hyperbolic tangent function into the controller, initial driving force is greatly reduced, and smooth start-up of the bridge crane can be guaranteed.

The invention relates to the field of relay protection of power systems, and discloses a DC side fault positioning system and method of a multi-end flexible DC power distribution system. The system includes a data acquiring module, a fault positioning module, a DC side fault protection module, and a result output module. The method herein includes the following steps: when a DC side line fault occurs, starting the low voltage over-current protection of a converter station, and transmitting a converter station locking signal; then using the phase jump characteristic of a hyperbolic tangent function to position a fault; locking respective bridge arm sub-module, based on the determination of whether a fault current stops attenuation, removing the DC fault current; after the disconnection of the fault line, closing the switch on non-fault lines. According to the invention, the method can position faults by using the characteristics of phase jump of the function when fault points and reference points match, which addresses the failure of conventional methods to start the closing of the switch by determining whether the corresponding DC line is energized. The method can also fast disconnect the fault lines and ensures normal operation of the non-fault lines.

A method for control that yaw of a vehicle on an expected path belongs to the control field of an unmanned vehicle. In order to solve the problem of the yaw control of the vehicle on the expected path, the main points are as follows: S2. The equivalent sliding film pendulum torque controller is established based on the vehicle dynamics model. The effect is that the pendulum torque controller is established based on the quasi-sliding film control, and the hyperbolic tangent function is used to replace the sign function, which effectively reduces the chattering phenomenon of the quasi-sliding film control.

The invention discloses an earthquake motion peak acceleration prediction method based on a second-order neuron deep neural network, belongs to the field of earthquake engineering, and aims to solve the technical problem of low precision of earthquake motion prediction. The earthquake motion peak acceleration prediction method comprises the following steps: 1, selecting earthquake magnitudes, projection distances, shear wave velocities, regions, cover layer thicknesses, fault types and periods as input parameters in a data set, and taking corresponding earthquake motion peak accelerations as target parameters; 2, establishing a deep neural network comprising three hidden layers, wherein neurons are second-order elements, a hyperbolic tangent function is adopted as an activation function, amean square error function and an Adam self-adaptive optimization function are adopted for back propagation, and an average absolute error function is adopted as an evaluation function; 3, training adeep neural network model; and 4, peak acceleration prediction. According to the method, a multi-input structure and a second-order neural network are adopted so that the precision of predicting theearthquake motion peak acceleration can be improved, and the applicability of the deep neural network model can be ensured.

The invention discloses an adaptive event trigger control method for a nonlinear uncertain system, which comprises the steps of firstly, for an input saturation phenomenon, adopting a smooth hyperbolic tangent function to approximate the saturation characteristic of system control input so as to avoid the phenomenon of unsmooth control input and facilitate the subsequent design; secondly, for an uncertain part of the system, improving the traditional Fuzzy Logic Systems (FLSs) to a certain extent, and taking the time-varying approximation error of an approximation system into consideration; finally, providing an event trigger control mechanism of a dynamic threshold for the bandwidth constraint existing in the actual system, wherein the trigger threshold changes along with the change of the control quantity at the last moment, so that more precise and flexible control can be realized, and the system bandwidth resources can be better saved on the basis of ensuring the control effect.

The invention discloses an integrated underwater thruster driving system and a control method, which realizes the integrated underwater thruster sensorless control. A linear back EMF sliding mode observer based on fuzzy switching gain adjustment is provided, The observer adopts a smooth hyperbolic tangent function and fuzzy rules to realize change of a gain value and reduce the chattering of the system, so that the control system can obtain smooth back EMF estimates without adding a low pass filter and a phase compensation module, thus avoiding the phase lag of the back EMF estimates. The lineback EMF estimates are corresponded to three virtual Hall signals and six discrete commutation signals are obtained directly without the need of fixed phase shift circuit and calculation of phase shift angle. the invention realizes the integrated underwater thruster sensorless control by estimating the line back EMF of the brushless DC motor loaded in the underwater thruster,.

The invention discloses a hyperbolic function sliding-mode control-based chaotic oscillation suppression method of a power system. The chaotic oscillation suppression method is implemented according to the following steps of 1, calculating a relative error e(t) according to a relative rotor angle actual value x<1>(t) and a relative rotor angular speed ideal value x<d>(t) of a second-order interconnection dual-motor power system; 2, setting a sliding-mode controller parameter; 3, calculating a motion state of a system sliding-mode surface according to the relative error e(t) and the sliding-mode controller parameter; and 4, generating a controller output u(t) by joint efforts of the sliding-mode controller parameter and a motion state equation, making the controller output acted on a rotor of a power generator, outputting the current relative rotor angle actual value x<1>(t), forwarding to the step 1, and adjusting the controller output according to the relative error. According to the hyperbolic function sliding-mode control-based chaotic oscillation suppression method of the power system, a chaotic oscillation phenomenon generated in the interconnection dual-motor power system is rapidly and effectively suppressed by the sliding-mode controller of a hyperbolic tangent function.

A driver modeling method for a horizontal / longitudinal motion unified framework includes the following steps: establishing a vehicle lateral single lane change trajectory analytic expression by usinga hyperbolic tangent function; acquiring road information, vehicle information, and the driving state information of the vehicle; obtaining the constraint of lateral displacement of the vehicle according to road state information; establishing an indicator function according to a driver characteristic and a safety requirement, obtaining an optimized lateral displacement parameter, and calculatingthe lateral force constraint of the vehicle in combination with the vehicle driving state and an environmental factor; obtaining the vehicle trajectory smoothness constraint by using the lateral forceconstraint; establishing an indicator function in combination with the driver characteristic to obtain an optimized value of a smoothness parameter; and finally obtaining an optimized trajectory expectation. With the lateral expectation and the longitudinal velocity as expected values, a driver model describing the horizontal / longitudinal motion unified framework is established to achieve a trajectory tracking purpose.

The invention discloses a gesture recognition method based on deep convolutional neural networks, and the method comprises the steps: (1) edge detection processing and sample set partition are performed on sample images of a training set; (2) deep convolutional neural networks are constructed; (3) an activation function and a loss function are determined; (4) deep neural networks are trained; (5) gesture recognition is realized according to the deep convolutional neural networks after training: wherein the steps comprises: a) a hand-type image is extracted from to-be-recognized gesture data; b) edge detection and size normalization processing are performed on the hand-type image; c) and the normalized hand-type image is inputted to the deep convolutional neural networks, a kind of the current gesture is judged according to an output value of an output layer. According to the invention, multiple downsampling technology is used to construct the deep convolutional neural networks, a hyperbolic tangent function is adopted to serve as an activation function to train neural network, the efficiency of gesture recognition can be improved, and accuracy of gesture recognition can be improved.

The invention discloses a line single-phase ground fault relay protection method capable of preventing distributed capacitive current and transitional resistance. The method comprises the following steps: a lumped parameter model is utilized for modeling at first; the fault location is calculated; then the physical properties of voltage and current transmission of a transmission line are described by adopting a long-line equation; the obtained fault location is utilized to calculate the offset current of the transmission line; further, the line fault impedance is calculated according to the hyperbolic tangent functional relationship between the voltage drop from the relay location of the transmission line to a single-phase ground fault point and the fault location; and determining whether an action trip signal is sent or not according to the judgment on the amplitude value of the line fault impedance. When the method provided by the invention is adopted for single-phase ground fault relay protection of an ultra / extra-high voltage AC transmission line, the circuit fault impedance from the relay location of the transmission line to the single-phase ground fault point can be accurately calculated, having no reference to the voltage, the distributed capacitive current, the transitional resistance and the load current of the ground fault point, and the action performance can meet the requirements for relay protection selectivity, reliability, sensitivity and quickness.

The present invention provides a third-order strict-feedback chaotic proportion projection synchronization method based on improved pole assignment. The method comprises the steps of: establishing a proportion projection synchronization error system according to a state equation of an isomorphic or isomerous third-order strict-feedback chaotic system; designing a linear sliding mode surface and animproved rapid power reaching law; combining a pole assignment method and a sliding-mode control method, designing a single sliding-mode controller to perform balance control of the proportion projection synchronization error system to form a closed-loop system, wherein the closed-loop system can achieve the proportion projection synchronization control of a drive system and a response system. Inorder to weaken the buffeting influence, a hyperbolic tangent function is employed to replace the symbolic function. In a condition of the modeling indeterminacy and external interference signals, the single sliding-mode controller is employed to achieve proportion projection synchronization control of the isomorphic or isomerous third-order strict-feedback chaos in different initial states. Theexperiment simulation result shows the effectiveness of the method, the speed of the proportion projection synchronization is fast, and the robustness is good.

The invention discloses an electric turntable position tracking control method based on disturbance upper bound estimation. A turntable servo system pitching frame is selected as a study object, meanwhile, uncertainty and uncertain nonlinearity of system parameters are considered at the same time, and an excellent turntable position tracking controller is designed; for the uncertainty of the system parameters, a good parameter adaptation law is designed, and system parameters can be estimated and converged, thereby achieving good model compensation; for the uncertain nonlinearity of the system parameters, a good disturbance upper bound estimation law is designed, and the maximum upper bound of the uncertain nonlinearity is estimated; and a disturbance compensation term in the controller adopts a hyperbolic tangent function, thereby avoiding buffeting caused by a traditional sign function. Further, the effectiveness of the controller is verified through contrast simulation results.

The invention discloses a method for achieving circuit phase-to-phase fault relay protection with hyperbolic tangent function amplitude characteristics. The method includes the steps of calculating a hyperbolic tangent function amplitude at the position of an electric transmission circuit phase-to-phase short circuit fault point through single-end fault phase-to-phase electrical quantity by adopting distribution parameter modeling, judging whether the hyperbolic tangent function amplitude at the position of the electric transmission circuit phase-to-phase short circuit fault point is smaller than a hyperbolic tangent function amplitude at the electric transmission circuit protection setting range position or not, and giving out a motion tripping signal with a protection device if the hyperbolic tangent function amplitude at the position of the electric transmission circuit phase-to-phase short circuit fault point is smaller than the hyperbolic tangent function amplitude at the electric transmission circuit protection setting range position. According to the method, electric transmission circuit phase-to-phase short circuit fault relay protection is achieved, influences of transition resistance, distributed capacitance, fault positions and load currents on the protection motion performance are overcome, and the method is suitable for extra-high voltage alternating current electric transmission circuit phase-to-phase short circuit fault relay protection.

The present invention discloses a method of matching accumulation type MOS varactor electrical properties, adopting capacitance Cap of formula computing accumulation type MOS varactor. Vgs is voltage applied between the grid electrode and source electrode of the accumulation type MOS varactor, and the grid electrode is taken as a variable voltage terminal, and the source electrode and a drain electrode are shorted, so as to form a reference voltage terminal; tanh() is a hyperbolic tangent function; c1=a+b*Wg+c*Lg+d*Wg*Lg, c2=e*Wg*Lg, wherein a, b, c, d and e are constant and range from 0 to 10; Wg and Lg are channel width and channel lengthof the accumulation type MOS varactor respectively; and c3 and c4 are constant and range from minus 2 to 2. The model has good fitting capability on various accumulation type MOS varactors with different channel lengths and different channel widths.

The invention discloses a water level prediction method based on an extended wavelet neural network model. The method determines the number of input items by analyzing the correlation between input and output, thereby improving the prediction accuracy. The method includes the following steps: Step 1: Calculate the correlation between the input xt, xt-1, xt-2, . . . , xt-n sequence and the output x. Step 2: According to the correlation, each sequence xt, xt-1, xt-2,...,xt-n, n≥1 is decomposed by 3 layers of wavelets using db5 as the mother wavelet, and each sequence is decomposed by wavelets to obtain the wavelet decomposition Coefficient normalization processing is used as the input of the network, and different extended wavelet neural network models are established through different values ​​of n; Step 3: Select the hyperbolic tangent function and linear function as the activation functions of the hidden layer and the output layer respectively, according to the mean square The size of the error (MSE) is used to select the appropriate number of hidden layer nodes, and the Levenberg-Marquardt (LM) optimization algorithm is used to train the network; Step 4: To test the different extended networks obtained, and select the most appropriate extended network neural network, Improve forecast accuracy.