Description Title of Invention: CONTROL DEVICE FOR ELEVATOR, ELEVATOR APPARATUS, AND METHOD OF OBTAINING ROTATION ANGLE ERROR OF ROTATION DETECTING UNIT FOR ELECTRIC MOTOR FOR ELEVATOR Technical Field [0001] The present invention relates to estimation of an angle error in a control device for an elevator, in particular, a control device having a periodic torque pulsation or speed pulsation due to a periodic angle error of a rotation sensor mounted to an electric motor constructing a hoisting machine. Background Art [0002] In a resolver device serving as a related-art rotation sensor, an error waveform of a resolver is composed of a frequency component specific to the resolver. Because of reproducibility, a positional error is calculated by referring to a detected angle signal. The positional error is differentiated to calculate a speed error signal. The speed error signal is subjected to Fourier transform to calculate a magnitude of a detection error for each of a plurality of divided components. The calculated detection errors are synthesized to generate an error wave signal obtained by restoring the detection error contained in the angle signal detected by the resolver. An angle detection signal of the resolver, which contains the detection error, is corrected by using the generated error waveform signal. Through the Fourier transform performed on the speed error obtained from the positional error, each individual detection signal of the resolver can be precisely calculated. Through correction of the detected angle signal with use of the obtained detection error, a precise angle signal can be obtained. Citation List Patent Literature [0003] [PTL1]JP 2012-145371 A Summary of invention Technical Problem [0004] Hitherto, in a case where the angle error is estimated by using a signal containing a pulsation generated due to a periodic angle error, when a frequency of 1 the periodic angle error and a natural frequency of a mechanical system connected as a load of an electric motor match, the angle error and the mechanical system cause resonance to temporarily change an amplitude and a phase of a speed pulsation generated due to the angle error, thereby losing reproducibility. Thus, with a method of a related-art device, it is difficult to estimate the angle error. Therefore, there is a possibility of obtaining an erroneous result of estimation. [0005] In particular, in a system, e.g., an elevator, in which the natural frequency of the mechanical system changes in accordance with a length of a rope, the frequency at which the resonance occurs momentarily varies, and hence it is not clear where the resonance occurs. Thus, there is a possibility that the angle error may be erroneously estimated. When specific specifications of the mechanical system of the elevator are obtained in advance, the natural frequency, a gain characteristic, and a phase characteristic of the elevator are calculated so that the angle error can be estimated at a speed and a car position at which the resonance does not occur. However, mechanical specifications of the elevator frequently differ depending on a building in many cases. With a method relying on the prior information, design time becomes enormously longer. Therefore, in a case where the mechanical system having the resonance is connected as the load of the electric motor, when the angle error is to be estimated by using the signal containing the pulsation generated due to the angle error, there is a problem in determination of whether or not the result of estimation of the angle error is correct. [0006] Further, in the related-art device, the positional error is calculated by referring to the detected angle signal. The positional error is differentiated to calculate the speed error signal. The speed error signal is then subjected to the Fourier transform to estimate the angle error. Here, when the angle error is estimated by using the speed signal, accuracy of estimation of an angle resolution is determined by a speed resolution of an angle detector or a speed detector. Thus, a quantization error is generated in the angle detector or the speed detector with a low speed resolution. Therefore, there is a problem in that sufficient estimation accuracy for the angle error cannot be obtained. [0007] The present invention has been made to solve the problems described above, and therefore has an object to provide a control device for an elevator and the like, which include a rotation detecting unit for an electric motor containing a periodic angle error, and are capable of obtaining a highly reliable result of estimation of an angle error without erroneously estimating the angle error even when the angle error and a mechanical system resonate with each other. 2 Solution to Problem [0008] According to one embodiment of the present invention, there are provided a control device for an elevator and the like, including: a current detector configured to detect a current flowing through an electric motor configured to generate power for raising and lowering a car inside a hoistway; a rotation detecting unit configured to detect a rotation angle of the electric motor; a frequency analyzing unit configured to output a component at a specific frequency obtained by frequency analysis of the current detected by the current detector; and an angle error estimating unit configured to estimate an amplitude and a phase of a periodic angle error determined uniquely in accordance with the rotation angle from the rotation detecting unit by using the component at the specific frequency so as to output the estimated amplitude and phase as an angle error estimation value, in which the angle error estimating unit is configured to perform control so as to perform a learning operation for operating the car over a specific interval, acquire a plurality of the components at the specific frequency, which are obtained by inputting the current detected during the learning operation to the frequency analyzing unit, in a continuous manner, calculate an evaluation value being a geometric quantity on a coordinate plane formed by a set number of continuous components at the specific frequency among the acquired plurality of the components at the specific frequency, and calculate the angle error estimation value and associate the evaluation value and the angle error estimation value with each other so as to select the angle error estimation value when the evaluation value becomes minimum. Advantageous Effects of Invention [0009] According to the present invention, the control device for an elevator and the like, which are capable of obtaining a highly reliable result of estimation of an angle error without erroneously estimating the angle error even when the angle error and a mechanical system resonate with each other, can be provided. Brief Description of Drawings [0010] FIG. 1 is a configuration diagram for illustrating an example of a control device for an elevator according to the present invention. FIG. 2 is a configuration diagram for illustrating an example of a configuration of an angle error estimating unit illustrated in FIG. 1. FIG. 3 is a graph for showing an example of frequency characteristics of a speed controller illustrated in FIG. 1. FIG. 4 is a graph for showing an example of a gain characteristic of a mechanical system of the elevator. FIG. 5 is a graph for showing an example of a phase characteristic of the mechanical system of the elevator. FIG. 6 is a graph for showing an example of a coordinate plane of a Fourier coefficient calculated by a frequency analyzing unit illustrated in FIG. 1. FIG. 7 is a flowchart for illustrating an example of an operation of a learning operation in a first embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. FIG. 8 is a graph for showing an example of an area surrounded by coordinates of the Fourier coefficients calculated by the angle error estimating unit illustrated in FIG. 1. FIG. 9 is a flowchart for illustrating an example of an operation of a learning operation in a second embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. FIG. 10 is a flowchart for illustrating an example of an operation of a learning operation in a third embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. FIG. 11 is a flowchart for illustrating an example of an operation of a learning operation in a fourth embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. FIG. 12 is a configuration diagram for illustrating an example of a configuration of the angle error estimating unit illustrated in FIG. 1. FIG. 13 is a graph for showing an example of a line segment length between the coordinates of the Fourier coefficients calculated by the angle error estimating unit illustrated in FIG. 1. FIG. 14 is a flowchart for illustrating an example of an operation of a learning operation in a fifth embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. FIG. 15 is a flowchart for illustrating an example of an operation of a learning operation in a sixth embodiment of the present invention, which is performed by the angle error estimating unit illustrated in FIG. 1. Description of Embodiments [0011] In a control device for an elevator according to one embodiment of the 4 present invention: a frequency analyzing unit is configured to calculate an amplitude and a phase of a specific frequency component by performing frequency analysis on a current detected by a current detector; and an angle error estimating unit is configured to estimate an angle error composed of a specific frequency component as an angle error estimation value by using the amplitude and the phase at the specific frequency component, which are computed by the frequency analyzing unit, perform a learning operation for operating a car over a specific interval when an angle detection error is estimated, store a plurality of results of computation of the amplitude and the phase at the specific frequency component during the learning operation, calculate an evaluation value being a geometric quantity on coordinates created by the amplitudes and the phases of the plurality of stored specific frequency components, and select the angle error estimation value when the evaluation value becomes minimum. Therefore, a highly reliable estimation value of the angle error can be obtained without erroneously estimating the angle error under effects of resonance. [0012] Now, a control device for an elevator and the like according to each of embodiments of the present invention are described with reference to the drawings. In each of the embodiments, the same or corresponding portions are denoted by the same reference symbols, and the overlapping description thereof is omitted. [0013] First Embodiment FIG. 1 is a configuration diagram for illustrating an example of a control device for an elevator according to the present invention. In FIG. 1, a car 4 and a counterweight 5 of the elevator are connected to each other by a hoisting rope 6 so as to be suspended from a sheave 3 in a well bucket style. The sheave 3 is coupled to an electric motor 1 being an electric motor for driving the car 4. The car 4 is raised and lowered by power of the electric motor 1. The electric motor 1 configured to raise and lower the car 4 is, for example, a permanent-magnet synchronous motor. [0014] A rotation detecting unit 2 configured to detect a rotation angle of the electric motor 1 or the sheave 3 is mounted on the same axis as those of the electric motor 1 and the sheave 3. For example, angle information being a rotation angle of the electric motor 1, which is output from the rotation detecting unit 2 constructed of a resolver, an encoder, a magnetic sensor, or the like contains a periodic error determined uniquely in accordance with the rotation angle of the electric motor 1. Herein, the periodic error determined uniquely in accordance with the rotation angle of the electric motor 1 means, for example, an error with reproducibility in accordance with the rotation angle, specifically, generated at the same rotation angle position during each revolution, such as a detection error of the resolver or pulse omission or irregularity in distance between pulses due to a slit defect in an optical encoder. [0015} Among a frequency analyzing unit 8, an angle error estimating unit 9, a speed computing unit 10, a speed command computing unit 11, a speed controller 12, a current controller 13, and subtractors SU1 to SU3, which are illustrated as functional blocks described later, at least the frequency analyzing unit 8, the angle error estimating unit 9, the speed computing unit 10, the speed command computing unit 11, and the subtractors SU1 to SU3 are constructed of, for example, a computer including a processor and a memory, and are each configured to perform processing in accordance with a program stored in the memory and various types of setting information required for the processing. Further, the speed controller 12 and the current controller 13 may be constructed of the computer described above in a similar manner. Further, a portion indicated as each of the functional blocks may also be constructed of a digital circuit configured to execute each own function. The same applies to FIG. 2. [0016] The speed command computing unit 11 is configured to compute and output a speed command value to the electric motor 1. Although not shown, the speed command computing unit 11 may include a position control system. Even when the speed command computing unit 11 includes the position control system, the present invention is applicable. [0017] A difference between the speed command value from the speed command computing unit 11 and a rotation speed of the electric motor 1, which is computed in the speed computing unit 10, is input to the speed controller 12 from the subtractor SU1, and the speed controller 12 computes and outputs a current command value to the electric motor 1. The speed controller 12 may be configured by using any control technique such as PI control and PD control. [0018] The speed computing unit 10 is configured to compute and output the rotation speed of the electric motor 1 based on a corrected rotation angle that is corrected and output from the subtractor SU2, which is a difference between the rotation angle of the electric motor 1 being output from the rotation detecting unit 2, and an angle error estimation value of a periodic error determined uniquely in accordance with the rotation angle of the electric motor 1, which is estimated by the angle error estimating unit 9. In the simplest configuration, the speed computing unit 10 computes the rotation speed by temporal differentiation of the rotation angle. Further, smoothing may be performed through a low-pass filter (not shown) so as to remove noise generated by the temporal differentiation. Still further, the speed computing unit 10 may compute the rotation speed of the electric motor 1 at each preset given time interval or may include a configuration for measuring time so as to compute the rotation speed at each preset given rotation angle. [0019] A difference between the current command value from the speed controller 12 and a phase current being output from the current detector 7 or an axis current of the electric motor 1, which is obtained by d-q axis transform of the phase current through coordinate transform (not shown) is input to the current controller 13 from the subtractor SU3, and the current controller 13 computes and outputs a voltage command for the electric motor 1. A control technique of the current controller 13 is not limited as in the case of the speed controller 12. [0020] The current detector 7 is configured to detect a current of the electric motor 1. For example, when the electric motor 1 is a three-phase electric motor, a three-phase current may be measured although a two-phase current is measured in many cases. Although the current detector 7 measures an output current of the power converter 14 in FIG. 1, the current detector 7 may measure a bus current of the power converter 14 so as to estimate each phase current as in a current measuring method with a one-shunt resistor. Even in this case, the present invention is not affected thereby. [0021] The power converter 14 is configured to convert a power supply voltage (not shown) into a preferable variable voltage variable frequency based on the voltage command from the current controller 13. The power converter 14 of the present invention is a power converter configured to convert an AC voltage into a DC voltage by a converter and then convert the DC voltage into an AC voltage by an inverter as in the case of a commonly commercially available inverter device, or a variable-voltage variable-frequency power converter including a power converting device configured to directly convert a DC voltage into an AC variable voltage variable current as in the case of a matrix converter. [0022] Further, the power converter 14 according to the present invention may include a coordinate transform function in addition to the above-mentioned inverter. Specifically, the power converter 14 is represented to include a coordinate transform function of converting a d-q axis voltage command value into a phase voltage or a line voltage so as to obtain a voltage in accordance with the commanded voltage command value when the voltage command is the d-q axis voltage command value. Even when a device or correction unit configured to correct dead time of the power converter 14 is provided, the present invention can be applied. [0023] The frequency analyzing unit 8 is configured to perform frequency analysis on a current being the phase current or the axis current detected by the current detector 7 so as to output an amplitude and a phase at a specific frequency. Here, it is desirable that the frequency analyzing unit 8 have a configuration capable of obtaining the amplitude and the phase at the specific frequency of the input signal as in the case of Fourier transform, discrete Fourier transform, Fourier series expansion, or fast Fourier transform. However, as in a case of a filter obtained by combining a notch filter and a bandpass filter, a specific frequency signal may be extracted, and an amplitude computation and a phase computation are performed on, for example, an output current of the band pass filter by an amplitude detecting unit and a phase detecting unit (not shown) so as to compute the amplitude and the phase at the specific frequency of the input signal. The filter used herein may be an electric one obtained by combining a resistor, a capacitor, a coil, and the like, or may be processing to be performed in a computer. In the following, the frequency analyzing unit 8 is described as being configured to perform the Fourier transform. [0024] The angle error estimating unit 9 uses a Fourier coefficient being output from the frequency analyzing unit 8 to estimate a periodic angle error contained in the rotation angle being output of the rotation detecting unit 2. The angle error estimating unit 9 prestores a transform expression for calculating an angle error by using the Fourier coefficient, which is obtained by the Fourier transform of the current of the current detector 7 using information of the rotation angle of the electric motor, in a memory so as to calculate an estimation value of the angle error from the current by using the transform expression. Further, the angle error estimating unit 9 is configured to calculate an area of a region surrounded by the coordinates of the Fourier coefficients based on the Fourier coefficients obtained by the Fourier transform of the current of the current detector 7, which is the output from the frequency analyzing unit 8, using the information of the rotation angle of the electric motor, as the evaluation value being the geometric quantity on the coordinate plane. A method of calculating the area surrounded by the coordinates of the Fourier coefficients and meaning thereof are described later. The angle error estimated by the angle error estimating unit 9 is composed of an error amplitude and an error phase described later. When calculating the error amplitude and the error phase as the estimation value of the angle error, the angle error estimating unit 9 computes and outputs a sine wave or cosine wave correction signal by using the error amplitude and the error phase so as to reproduce the periodic angle error. [0025] FIG. 2 is a configuration diagram for illustrating an example of the angle error estimating unit 9. A learning speed computing unit 91 calculates the rotation speed of the electric motor 1 based on the rotation angle of the electric motor 1, which is detected by the rotation detecting unit 2. In the simplest configuration, the learning speed computing unit 91 computes the rotation speed by temporal differentiation of the rotation angle. Further, smoothing may be performed through a low-pass filter so as to remove noise generated by the temporal differentiation. Still further, the learning-speed computing unit 91 may compute the rotation speed of the electric motor 1 at each preset given time interval or may include a configuration for measuring time so as to compute the rotation speed at each preset given rotation angle. The rotation angle detected by the rotation detecting unit 2 contains the periodic angle error. Accordingly, the rotation speed of the electric motor 1, which is calculated by the learning speed computing unit 91, contains periodic speed pulsations. Speed information required in the angle error estimating unit 9 is used so as to determine whether or not the rotation speed of the electric motor 1 has reached a preset speed to achieve a constant speed running state based on a set speed during an angle error learning operation described later. Therefore, even when the pulsation is contained in the speed information, the determination for the achievement of the constant speed running state can be made without any problem. [0026] A car position computing unit 92 is configured to calculate and output a position of the car 4 inside a hoistway based on the rotation angle of the electric motor 1, which is the output from the rotation detecting unit 2. A reference position inside the hoistway may be a bottom floor or a top floor, or may be based on an arbitrary floor as a reference. The rotation angle of the electric motor 1, which is the output from the rotation detecting unit 2, contains a periodic angle error. Therefore, the car position calculated by the car position computing unit 92 also contains an error. Positional information of the car 4, which is necessary in the present invention, is used in an angle error learning operation described later to determine whether or not running over a specific interval has been completed. Therefore, even when the error is contained in the positional information of the car 4, the determination for the completion of running over the specific interval can be made without any problem. Further, the car position computing unit 92 may count, for example, the number of times of detection of a door zone plate so as to determine the completion of running over the specific interval without calculating the position of the car 4 based on the rotation angle. Further, the completion of running over the specific zone may be determined by a position switch that is provided in the hoistway and is configured to notify the reference position, e.g., the top fioor or the bottom floor. [0027] A learning determining unit 93 determines whether or not the rotation speed of the electric motor 1, which is the output of the learning speed computing unit 91, has reached the constant speed running state and whether or not the car 4 is running in a preset specific interval based on the position of the car 4, which is the output of the car position computing unit 92. The learning determining unit 93 outputs a learning command while the rotation speed is in the constant speed running state and the car is running in the specific interval, and otherwise does not output the learning command. Specifically, the angle error estimating unit 9 executes estimation of the angle error when the rotation speed of the electric motor 1 is constant and a frequency of the angle error is constant. As a result, the frequency of the angle error can be treated as being known. [0028] When receiving the leaning command from the learning determining unit 93, an angle error computing unit 94 computes the angle error by using the Fourier coefficient being the output of the frequency analyzing unit 8, which is obtained by the Fourier transform of the current of the current detecting unit 7 using the information of the rotation angle of the electric motor 1. The angle error computing unit 94 prestores a transform expression for obtaining the angle error from the Fourier coefficient corresponding to a result of the Fourier transform of the current using the information of the rotation angle of the electric motor 1 in a memory so as to compute the angle error from the Fourier coefficient. The angle error calculated by the angle error computing unit 94 is composed of the error amplitude and the error phase described later. [0029] An area computing unit 95 calculates an area of a region formed by coordinates of the Fourier coefficients based on the Fourier coefficients being the output of the frequency analyzing unit 8, which is obtained by the Fourier transform of the current of the current detector 7 using the information of the rotation angle of the electric motor 1. The area of the region formed by the coordinates of the Fourier coefficients is output to an output determining unit 96 so as to be used for the determination of successful/unsuccessful estimation of the angle error. The area created by the coordinates of the Fourier coefficients and meaning thereof are described later. [0030] The output determining unit 96 selects and outputs the error amplitude and the error phase when the area formed by the coordinates of the Fourier coefficients being the output of the area computing unit 95 becomes minimum. The minimum area formed by the coordinates of the Fourier coefficient is equivalent to the smallest effect of resonance. Therefore, the estimation value of the angle error with the small effect of resonance can be selected. [0031] An error signal computing unit 97 calculates and outputs a correction signal (angle error estimation value) for correcting the periodic angle error of the rotation detecting unit 2 by using the error amplitude and the error phase, which are the output of the output determining unit 96. The correction signal is a value obtained by multiplying a sine value or a cosine value of the rotation angle, which is obtained by adding the error phase calculated in the angle error computing unit 94 to the rotation angle of the electric motor 1 being the output of the rotation detecting unit 2, by the error amplitude (Expression (1) described later). [0032] Next, the periodic angle error contained in the angle output from the rotation detecting unit 2 is described. The periodic angle error of the rotation detecting unit 2 can be approximately represented by using the sine wave as expressed by Expression (1). There is no essential difference between a description with the sine wave and a description with the cosine wave. Therefore, the description is uniquely given with the sine wave in the present invention. [0033] ee=Atsin(Xem+<|>) (1) [0034] 0e: periodic angle error of the rotation detecting unit 2 X: order (known value) of the angle error of the rotation detecting unit 2 with respect to a mechanical angle of the electric motor 1 Gm: rotation angle of the electric motor 1 A-i: error amplitude of the angle error of the rotation detecting unit 2 0: phase shift (error phase) of the rotation detecting unit 2 with respect to the mechanical angle of the electric motor 1 [0035] X denotes the order of the angle error of the rotation detecting unit 2 with respect to the mechanical angle of the electric motor 1, and is a known value. Thus, when the rotation angle 9m of the electric motor 1, specifically, the rotation speed of the electric motor 1 is known, the frequency of the periodic angle error of the rotation detecting unit 2, which is expressed by Expression (1), can be obtained. Further, Ai denotes the error amplitude of the angle error of the rotation detecting unit 2, and (j> denotes the phase shift (error phase) of the rotation detecting unit 2 with respect to the mechanical angle of the electric motor 1. [0036] In the angle error estimating unit 9, the correction signal expressed by Expression (1) is calculated by using a result of estimation of the error amplitude A-i and a result of estimation of the error phase <\>. [0037] Next, the periodic angle error expressed by Expression (1) is converted into a periodic speed error by the speed computing unit 10 as expressed by Expression (2). [0038] aje=XA1cocos(Xem+4))=A2COs(Xem+c})) (2) ioe: periodic speed error of the rotation detecting unit 2 A2: amplitude of the speed error based on the angle error of Expression (1) w: rotation speed of the electric motor [0039] Thus, the rotation speed of the electric motor 1, which is output from the speed computing unit 10, contains the periodic speed error expressed by Expression (2). Then, the speed output from the speed computing unit 10 is compared with the speed command value output from the speed command computing unit 11 so as to be input to the speed controller 12. The current command is determined from a difference between the speed command value and the detected speed in the speed controller 12. However, the rotation speed output from the speed computing unit 10 contains the periodic speed error as expressed by Expression (2). Thus, the current command calculated by the speed controller 12 contains a pulsation generated due to Expression (2), specifically, a pulsation generated due to the angle error of the rotation detecting unit 2 expressed by Expression (1). The pulsation of the current command, which is calculated by the speed controller 12, is expressed by Expression (3) from Expression (2). [0040] le=A3cos(Xem+(|)+(t)c) (3) le: pulsation of the current command A3: amplitude of the current pulsation due to the angle error Oc: phase delay generated by the speed controller 12 [0041] The current pulsation expressed in Expression (3) is expressed by Expression (4) through Fourier series expansion. [0042] le-Ancos(X9m)+Bnsin(Xem) (4) [0043] The current command output from the speed controller 12 contains the current pulsation expressed by Expression (4). Therefore, the current detected by the current detector 7 also contains the current pulsation expressed by Expression (4). Through synthesis of a trigonometric function, the current pulsation expressed by Expression (4) can be rewritten as follows. [0044] le=V(An2+Bn2)-sin(X9m+Y) (5) G=V(An2+Bn2) Y=tan'1(An/Bn) [0045] In Expressions (5): G=V(An2-*-Bn2) expresses an amplitude of the current pulsation generated due to the angle error of the rotation detecting unit 2, and G=V(An2+Bn2)=A3; and Y=tan"1(An/Bn) expresses a phase difference of the current pulsation from the mechanical angle of the electric motor 1, and Y=tan~1(An/Bn)=c()+<|)c. In the following description, G=V(An2+Bn2) is referred to as the amplitude of the current pulsation, and Y=tan"1(An/Bn) is referred to as the phase of the current pulsation. [0046] Now, a method of obtaining the error amplitude A-i and the error phase tj) of the angle error from Fourier coefficients An and Bn obtained from the result of frequency analysis of the current is described. First, the error phase $ of the angle error can be obtained as Expression (6). [0047] c (6) [0048] The phase delay Regenerated by the speed controller 12 is determined by the frequency characteristics of the speed controller 12. A frequency of the angle error is known, and therefore a frequency of the speed error is a known value. FIG. 3 is a graph for showing frequency characteristics of a gain and a phase of the speed controller 12. For example, when the frequency of the angle error is A, the phase delay generated by the speed controller 12 is -150 [deg]. Further, when the frequency of the angle error is B, the phase delay generated by the speed controller 12 is -170 [deg]. The phase delay of the speed controller 12 is determined uniquely by the speed controller 12. Thus, from the frequency characteristics of the speed controller 12, the phase delay $c generated by the speed controller 12 can be obtained. Thus, the error phase can be obtained by subtracting the phase delay $c generated by the speed controller 12. For the amplitude, V(An2+Bn2) in Expressions (7) can be directly detected. Thus, the amplitude Ai of the angle error can be obtained in the same procedure as that of Expressions (7). [0056] Next, description is made of a learning operation method for selecting the estimation value when changes in the amplitude and the phase of the current pulsation are small, specifically, the frequency of the angle error of the rotation detecting unit 2 and a resonant frequency of the mechanical system of the elevator do not match to make the effects of resonance smaller in the estimation of the angle error expressed by Expression (1) according to this embodiment. [0057] First, characteristics of the mechanical system of the elevator are described. FIG. 4 and FIG. 5 are graphs for showing an example of a gain characteristic and an example of a phase characteristic, respectively, from the angle error contained in the rotation angle of the electric motor 1, which is detected by the rotation detecting unit 2, to the current detected by the current detector 7. As shown in FIG. 4 and FIG. 5, when the frequency of the angle error is A and C, the gain and the phase have constant values regardless of the position of the car 4 inside the hoistway. On the other hand, when the frequency of the angle error is B, the gain and the phase are the same in a case where the position of the car 4 is in the vicinity of the bottom floor indicated by the broken line and in the vicinity of the top floor indicated by the solid line. In the vicinity of a middle floor indicated by the dot line, however, resonance characteristics are exhibited. The characteristics shown in FIG. 4 and FIG. 5 are one example. The elevator has different mechanical specifications for each building. Thus, the gain characteristic and the phase characteristic shown in FIG. 4 and FIG. 5 differ for each building including the elevator. Further, a speed of the elevator also differs for each building. Therefore, it is difficult to know in advance at which position and speed the angle error and the mechanical system resonate with each other. Therefore, it is difficult to prevent the resonance based on the prior information. Thus, it is desirable that the learning of the angle error and the determination of successful/unsuccessful estimation be performed simultaneously. [0058] Next, a method of determining the successful/unsuccessful estimation by the angle error estimating unit 9 is described. F!G. 6 is a graph for showing the two Fourier coefficients An and Bn calculated by the frequency analyzing unit 8, which are plotted with the horizontal angle indicating Bn and the vertical axis indicating An, when the car 4 of the elevator 4 is made to run over the specific interval. The plane of FIG. 6 is hereinafter referred to as "Fourier coefficient coordinate plane". As is understood from Expressions (5), a distance G=V(An2+Bn2) from a point of origin in FIG. 6 is equal to the amplitude of the current pulsation, whereas an angle Y=tan"1(An/Bn) formed by a vector having the distance from the point of origin is equal to the phase of the current pulsation. [0059] Next, a relationship between the area surrounded by the coordinates of the Fourier coefficients, the amplitude of the current pulsation, and the phase of the current pulsation on the Fourier coefficient coordinate plane is described. Considering changes in the amplitude G=V(An2+Bn2) of the current pulsation and the phase y=tan~1(An/Bn)of the current pulsation on the Fourier coefficient coordinate plane of FIG. 6, the area surrounded by the coordinates of three or more Fourier coefficients becomes smaller when the changes in the amplitude G=V(An2+Bn2) of the current pulsation and the phase Y-tan'1(An/Bn) of the current pulsation are small. In a section A of FIG. 6, the area surrounded by the coordinates of the Fourier coefficients is small because of small changes of the coordinates of the Fourier coefficients. Therefore, the changes in the amplitude G=V(An2+Bn2) of the current pulsation and the phase Y-tan"1(An/Bn)of the current pulsation are small. Further, in a section B of FIG. 6, the area surrounded by the coordinates of the Fourier coefficients is large, and the changes in the amplitude G=V(An2+Bn2) of the current pulsation and the phase Y=tan"1(An/Bn)of the current pulsation are large as compared to the section A. Thus, the calculation of the area of the coordinates of the Fourier coefficients on the Fourier coefficient coordinate plane is equivalent to calculation of change amounts in the amplitude and the phase of the current pulsation. [0060] On the Fourier coefficient coordinate plane, the area surrounded by the coordinates of the three or more Fourier coefficients can be calculated by a coordinate method. For example, assuming that the Fourier coefficients calculated by the frequency analyzing unit 8 when the car 4 of the elevator runs over the specific interval are (Bni, Ani), (Bn2> An2). and (Bn3, An3) in chronological order, an area S surrounded by those three sets of the coordinates of the Fourier coefficients is calculated by Expression (9). [0061] S=(1/2)|(BniAn2-Bn2Ani)+(Bn2An3-Bn3An2)+(Bn3Anl-BnlAn3)| (9) [0062] When a distance V(Ani2+Bni2) from the point of origin to the coordinates (Bni, Ani)isGi, a distance V(An22+Bn22) from the point of origin to the coordinates (Bn2, An2) is G2, and a distance V(An32+Bn32) from the point of origin to the coordinates (Bn3, An3) is G3, specifically, the amplitudes of the current pulsations at the respective coordinates are Gi, G2, and G3, and when an angle tan"1(Ani/Bn1) formed by a distance vector from the point of origin to the coordinates (Bn1, Ani) is yi, an angle tan"1(An2/Bn2) formed by a distance vector from the point of origin to the coordinates (Bn2> An2) is Y2, and an angle tan"1(An3/Bn3) formed by a distance vector from the point of origin to the coordinates (Bn3, An3) is y3, specifically, the phases of the current pulsations at the respective coordinates are Yt, Y2, and Y3> the area expressed by Expression (9) can be rewritten as follows. [0063] S=(1/2)|G1G2sin(Y2-Yi)+G2G3Sin(Y3-Y2)+G1G3Sin(Yi-Y3)[ (10) [0064] The gain characteristic and the phase characteristic from the angle error contained in the rotation angle of the electric motor 1, which is detected by the rotation detecting unit 2, to the current detected by the current detector 7, are the characteristics as shown in FIG. 4 and FIG. 5. For example, when the frequency of the angle error is A or C, the gain and the phase have constant values regardless of the position of the car 4 inside the hoistway. Specifically, when the frequency of the angle error is A or C, the resonance with the mechanical system of the elevator does not occur. Therefore, G-i, G2l and G3 and Yi> Y2. and Y3 are constant values. As a result, the area calculated by Expression (9) becomes zero. On the other hand, in a case where the frequency of the angle error is B, the gain, specifically, the amplitude, and the phase are constant values when the car 4 is in the vicinity of the bottom floor and the top floor. However, when the car 4 comes closer to the middle floor, the gain and the phase greatly change. Specifically, in a case where the frequency of the angle error is B, the resonance occurs in the vicinity of the middle ffoor. In other locations, the resonance does not occur. In this case, the area calculated by Expression (9) becomes zero in the vicinity of the bottom floor and the top floor. However, the area expressed by Expression (9) increases as the car 4 comes closer to the middle floor. When the resonance occurs, the gain and the phase change simultaneously. Therefore, through calculation of the area expressed by Expression (9), the changes in the amplitude and the phase of the current pulsation can be calculated. [0065] Further, even when the frequency analyzing unit 8 is the filter obtained by combining the notch filter and the bandpass filter to extract the specific frequency signal such that the amplitude and the phase at the specific frequency of the input signal are computed by the amplitude detecting unit and the phase detecting unit, the area can be calculated in the same procedure. Specifically, when the frequency analyzing unit 8 is the filter obtained by combining the notch filter and the bandpass filter to extract the specific frequency signal such that the amplitude and the phase at the specific frequency of the input signal are computed by the amplitude detecting unit and the phase detecting unit, the amplitude G of the current pulsation and the phase y of the current pulsation in Expression (10) are obtained. Thus, the area can be calculated by Expression (10). Further, when the amplitude G and the error phase § of the current pulsation are used, An=Gsin(<|)) and Bn=Gcos(