5D Library › 五节点循环网络动力学、黄经历表与多尺度振子扰动 · English
第一章 开普勒方程与均时差:钟表时间、太阳时间与节气的日期
两种「中午」
钟表上的十二点与太阳最高的时刻并不重合。原因有两个:标准时区以一条经线为准,而大多数地方不在那条经线上;即使在那条经线上,真太阳在一年中也时快时慢。后者称为均时差(equation of time),全年在大约 −14 分钟到 +16 分钟之间摆动。本章用开普勒方程把它从头算出来,再问一个离散的问题:当一个节气的时刻落在午夜附近时,按钟表时间与按当地太阳时间,它会不会被记在不同的日子?
开普勒方程与牛顿法
推导
开普勒方程。 地球绕日轨道的平近点角 随时间均匀增长,偏近点角 满足 为轨道偏心率。这个方程没有初等闭式解。牛顿法从 出发迭代 在根附近误差每步大致平方一次(二次收敛)(Danby 1988)。
推导
# Chapter 1. p01: Kepler's equation M = E - e sin E, solved by Newton's
# method. Newton: E <- E - (E - e sin E - M) / (1 - e cos E), started at E = M
# (or at pi for e > 0.8, the standard safe start). The program counts
# iterations to a residual below 1e-15 rad over 3,600 mean anomalies, for
# Earth's eccentricity and for larger ones, and reports the largest gap
# between E and M - the "equation of centre" in eccentric anomaly.
import math
def solve(m, e, tol=1e-15, cap=60):
x = math.pi if e > 0.8 else m
for k in range(1, cap + 1):
f = x - e * math.sin(x) - m
x -= f / (1 - e * math.cos(x))
if abs(x - e * math.sin(x) - m) < tol:
return x, k
return x, cap
if __name__ == "__main__":
grid = [2 * math.pi * i / 3600 for i in range(3600)]
print(" e max iterations mean iterations max |E - M| (deg)")
for e in (0.0167, 0.1, 0.3, 0.6, 0.9, 0.99):
its, gap = [], 0.0
for m in grid:
x, k = solve(m, e)
its.append(k)
gap = max(gap, abs(x - m))
print(" {:6.4f} {:3d} {:5.2f} {:8.4f}"
.format(e, max(its), sum(its) / len(its), math.degrees(gap)))
e = 0.0167
m = math.radians(60.0)
x = m
print("\nEarth, M = 60 deg, step by step:")
for k in range(4):
x -= (x - e * math.sin(x) - m) / (1 - e * math.cos(x))
print(" step {}: E = {:.15f} residual {:.1e}".format(
k + 1, x, abs(x - e * math.sin(x) - m)))
print("\nReading: for Earth's orbit Newton's method needs at most three")
print("steps to machine precision - the residual squares each step -")
print("and E never strays more than about one degree from M. The")
print("eccentricity is small, which is why the equation of time is a")
print("few minutes, not hours.")
# e max iterations mean iterations max |E - M| (deg)
# 0.0167 3 2.60 0.9568
# 0.1000 3 2.94 5.7296
# 0.3000 4 3.67 17.1887
# 0.6000 5 4.29 34.3775
# 0.9000 8 4.65 51.5662
# 0.9900 10 4.62 56.7228
#
# Earth, M = 60 deg, step by step:
# step 1: E = 1.061781955213338 residual 1.5e-06
# step 2: E = 1.061780400132823 residual 1.8e-14
# step 3: E = 1.061780400132805 residual 0.0e+00
# step 4: E = 1.061780400132805 residual 0.0e+00
#
# Reading: for Earth's orbit Newton's method needs at most three
# steps to machine precision - the residual squares each step -
# and E never strays more than about one degree from M. The
# eccentricity is small, which is why the equation of time is a
# few minutes, not hours.对地球的偏心率 0.0167,3,600 个均匀分布的平近点角里,牛顿法最多三步就到机器精度,平均 2.60 步; 与 之差从不超过 0.9568 度。逐步输出显示了二次收敛:第一步残差 1.5e-06,第二步已是 1.8e-14。偏心率增大时迭代次数随之增加,到 0.99 时最多 10 步,但仍然收敛。地球的偏心率很小,这正是均时差只有十几分钟而不是几个小时的原因。
均时差的两个来源
推导
从轨道到赤经。 由 得真近点角 ,太阳的黄经 ( 为近地点黄经),再按黄赤交角 换成赤经 平太阳的黄经为 。均时差为 ,每度折合 4 分钟(视太阳时减平太阳时)(Meeus 1998; Hughes et al. 1989)。只保留偏心率(令 )得到一年一周的分量,只保留黄赤交角(令 )得到一年两周的分量。
推导
# Chapter 1. p02: The equation of time from Kepler's equation and the
# obliquity.
# Elements of the Sun's apparent orbit, d = days from 2000 Jan 1.5 TT
# (a low-precision set of the kind tabulated in astronomical almanacs):
# mean anomaly M = 357.5291 + 0.98560028 d (deg)
# perigee longitude w = 282.9404 + 4.70935e-5 d
# eccentricity e = 0.016709, obliquity eps = 23.4393 - 3.563e-7 d
# Mean longitude L = M + w. Kepler's equation gives E, then the true
# anomaly v and ecliptic longitude lam = v + w; right ascension
# alpha = atan2(cos eps sin lam, cos lam). Equation of time
# = L - alpha, at 4 minutes per degree (apparent minus mean solar time).
# Year 2027, noon UT each day. Parts: e only (eps = 0), eps only (e = 0).
import datetime
import math
def kepler(m, e):
x = m
for _ in range(8):
x -= (x - e * math.sin(x) - m) / (1 - e * math.cos(x))
return x
def eot(d, use_e=True, use_eps=True):
m = math.radians((357.5291 + 0.98560028 * d) % 360)
w = math.radians(282.9404 + 4.70935e-5 * d)
e = 0.016709 if use_e else 0.0
eps = math.radians(23.4393 - 3.563e-7 * d) if use_eps else 0.0
ea = kepler(m, e)
v = 2 * math.atan2(math.sqrt(1 + e) * math.sin(ea / 2),
math.sqrt(1 - e) * math.cos(ea / 2))
lam = v + w
alpha = math.atan2(math.cos(eps) * math.sin(lam), math.cos(lam))
diff = math.degrees(m + w - alpha)
diff = (diff + 180) % 360 - 180
return 4 * diff
if __name__ == "__main__":
base = datetime.date(2000, 1, 1)
days = [datetime.date(2027, 1, 1) + datetime.timedelta(i)
for i in range(365)]
rows = [(x, eot((x - base).days)) for x in days]
lo = min(rows, key=lambda r: r[1])
hi = max(rows, key=lambda r: r[1])
print("2027: minimum {:+.1f} min on {}, maximum {:+.1f} min on {}"
.format(lo[1], lo[0], hi[1], hi[0]))
print("\nlocal extremes and zero crossings:")
for (a, fa), (b, fb), (c, fc) in zip(rows, rows[1:], rows[2:]):
if (fb - fa) * (fc - fb) < 0:
print(" {} extreme {:+6.2f} min".format(b, fb))
if fa * fb < 0:
print(" {} crosses zero".format(b))
print("\n month total e only eps only (min, on the 1st)")
for mth in range(1, 13):
d = (datetime.date(2027, mth, 1) - base).days
print(" {:5d} {:+6.2f} {:+6.2f} {:+7.2f}".format(
mth, eot(d), eot(d, use_eps=False), eot(d, use_e=False)))
amp_e = max(abs(eot((x - base).days, use_eps=False)) for x in days)
amp_o = max(abs(eot((x - base).days, use_e=False)) for x in days)
print("\namplitudes: eccentricity part {:.2f} min, obliquity part"
" {:.2f} min".format(amp_e, amp_o))
print("\nReading: the total runs from about -14.2 to +16.4 minutes. It")
print("is the sum of a one-a-year wave from the elliptical orbit and a")
print("twice-a-year wave from the tilted axis; neither alone reaches")
print("the extremes, and the sum crosses zero four times.")
# 2027: minimum -14.2 min on 2027-02-11, maximum +16.4 min on 2027-11-03
#
# local extremes and zero crossings:
# 2027-02-11 extreme -14.20 min
# 2027-04-16 crosses zero
# 2027-05-14 extreme +3.64 min
# 2027-06-13 crosses zero
# 2027-07-26 extreme -6.57 min
# 2027-09-02 crosses zero
# 2027-11-03 extreme +16.45 min
# 2027-12-26 crosses zero
#
# month total e only eps only (min, on the 1st)
# 1 -3.45 +0.34 -3.82
# 2 -13.52 -3.67 -9.83
# 3 -12.36 -6.40 -6.37
# 4 -3.93 -7.65 +3.13
# 5 +2.86 -6.83 +9.57
# 6 +2.16 -4.17 +6.62
# 7 -3.87 -0.53 -3.29
# 8 -6.38 +3.35 -9.77
# 9 -0.08 +6.35 -6.02
# 10 +10.27 +7.63 +3.24
# 11 +16.42 +6.88 +9.67
# 12 +11.08 +4.28 +6.52
#
# amplitudes: eccentricity part 7.66 min, obliquity part 9.86 min
#
# Reading: the total runs from about -14.2 to +16.4 minutes. It
# is the sum of a one-a-year wave from the elliptical orbit and a
# twice-a-year wave from the tilted axis; neither alone reaches
# the extremes, and the sum crosses zero four times.2027 年的均时差最小值是 −14.2 分钟(2 月 11 日),最大值是 +16.4 分钟(11 月 3 日),中间还有 5 月 14 日的 +3.64 与 7 月 26 日的 −6.57 两个次极值,全年四次过零。偏心率分量的幅度是 7.66 分钟,黄赤交角分量是 9.86 分钟;两者单独都达不到总曲线的极值,极值来自两条波在 2 月与 11 月的同相叠加。
节气的日期:一个离散化的问题
节气以太阳的视黄经为准:黄经每到 15 度的整数倍,就开始一个新的节气。这是一个时刻,不是一个日子;把它记到哪一天,要看用哪一种钟。第三段程序算出 2001 到 2100 年的全部 2,400 个节气时刻,分别按东八区标准时与当地视太阳时(世界时 + 经度/15 小时 + 均时差)记日,统计两种记法日期不同的比例。
推导
# Chapter 1. p03: Clock dates versus solar dates for the 24 solar terms.
# A solar term (jieqi) begins when the Sun's apparent ecliptic longitude
# reaches a multiple of 15 deg. Its civil date depends on which clock
# dates it. Instants for 2001-2100 come from the low-precision solar
# theory below (good to about 0.01 deg, i.e. about 15 minutes in time;
# the 69 s difference between TT and UT is ignored at that precision).
# Each instant is dated two ways: China Standard Time (UTC+8) and local
# apparent solar time (UT + longitude/15 h + equation of time). The
# share of the 2,400 instants whose two dates differ is reported.
import math
def sun_lon(jd):
t = (jd - 2451545.0) / 36525
l0 = 280.46646 + 36000.76983 * t + 0.0003032 * t * t
m = math.radians(357.52911 + 35999.05029 * t - 0.0001537 * t * t)
c = ((1.914602 - 0.004817 * t - 0.000014 * t * t) * math.sin(m)
+ (0.019993 - 0.000101 * t) * math.sin(2 * m)
+ 0.000289 * math.sin(3 * m))
om = math.radians(125.04 - 1934.136 * t)
return (l0 + c - 0.00569 - 0.00478 * math.sin(om)) % 360
def eot_min(jd):
t = (jd - 2451545.0) / 36525
l0 = math.radians(280.46646 + 36000.76983 * t)
e = 0.016708634
m = math.radians(357.52911 + 35999.05029 * t)
y = math.tan(math.radians(23.4393) / 2) ** 2
val = (y * math.sin(2 * l0) - 2 * e * math.sin(m)
+ 4 * e * y * math.sin(m) * math.cos(2 * l0)
- 0.5 * y * y * math.sin(4 * l0) - 1.25 * e * e * math.sin(2 * m))
return 4 * math.degrees(val)
def instants(y0=2001, y1=2100):
out = []
jd = 2451910.5 # 2001 Jan 1
end = jd + 365.2422 * (y1 - y0 + 1)
while jd < end:
a, b = sun_lon(jd), sun_lon(jd + 1)
k = math.floor(a / 15) + 1
target = (15 * k) % 360
if (b - target) % 360 < (b - a) % 360:
lo, hi = jd, jd + 1
for _ in range(50):
mid = (lo + hi) / 2
if (sun_lon(mid) - target) % 360 < 180:
hi = mid
else:
lo = mid
out.append((lo + hi) / 2)
jd += 1
return out
def day_of(jd_local):
return math.floor(jd_local + 0.5)
if __name__ == "__main__":
ins = instants()
print("solar-term instants found, 2001-2100: {}".format(len(ins)))
eq = min(ins, key=lambda j: abs(j - 2460389.6)) # March 2024
h = ((eq + 0.5) % 1) * 24
print("check: March equinox 2024 at {:02d}:{:02d} UT by this theory"
" (published: 03:06 UT)".format(int(h), int(round(h % 1 * 60))))
print("\n place longitude dates differ share")
for name, lon in (("zone meridian", 120.0), ("Beijing", 116.4),
("Chengdu", 104.1), ("Lhasa", 91.1),
("Urumqi", 87.6)):
diff = 0
for jd in ins:
cst = day_of(jd + 8 / 24)
lat = day_of(jd + lon / 360 + eot_min(jd) / 1440)
diff += cst != lat
print(" {:<13} {:7.1f}E {:6d} {:6.2%}".format(
name, lon, diff, diff / len(ins)))
near = sum(1 for jd in ins
if abs(((jd + 8 / 24 + 0.5) % 1) * 1440) < 15
or abs(((jd + 8 / 24 + 0.5) % 1) * 1440 - 1440) < 15)
print("\ninstants within 15 minutes of midnight (UTC+8): {} ({:.2%})"
.format(near, near / len(ins)))
print("\nReading: on the zone meridian only the equation of time")
print("separates the two clocks, and about one solar term in two")
print("hundred changes date; at Beijing about one in a hundred. In")
print("the far west of the zone the clock runs more than two hours")
print("ahead of the Sun and the share approaches one in ten.")
print("A few dozen instants sit so near midnight that this ephemeris")
print("cannot date them at all.")
# solar-term instants found, 2001-2100: 2400
# check: March equinox 2024 at 03:04 UT by this theory (published: 03:06 UT)
#
# place longitude dates differ share
# zone meridian 120.0E 13 0.54%
# Beijing 116.4E 26 1.08%
# Chengdu 104.1E 109 4.54%
# Lhasa 91.1E 198 8.25%
# Urumqi 87.6E 220 9.17%
#
# instants within 15 minutes of midnight (UTC+8): 56 (2.33%)
#
# Reading: on the zone meridian only the equation of time
# separates the two clocks, and about one solar term in two
# hundred changes date; at Beijing about one in a hundred. In
# the far west of the zone the clock runs more than two hours
# ahead of the Sun and the share approaches one in ten.
# A few dozen instants sit so near midnight that this ephemeris
# cannot date them at all.程序先自检:它算出的 2024 年春分是世界时 03:04,公布值为 03:06,相差两分钟,与这套低精度太阳理论约 0.01 度(约 15 分钟)的精度相符。在时区的基准经线(东经 120 度)上,两种钟只差均时差,2,400 个节气中有 13 个(0.54%)日期不同;北京(东经 116.4 度)是 26 个(1.08%)。越往西,钟表越领先太阳:成都 4.54%,拉萨 8.25%,乌鲁木齐 9.17%——接近十分之一。
还有 56 个时刻(2.33%)落在东八区午夜前后 15 分钟以内。以本章所用的历表精度,这些节气属于哪一天无法判定;判定它们需要更精确的历表,第八章给出这项核对的预注册方案。
注
这一章说明了什么。 「节气在哪一天」不是一个天文事实,而是一个天文时刻加上一个记时约定。约定不同,约十分之一的节气会换一天。任何依据节气日期做的计算——历法、农事表、程序里的日期查表——都隐含了一个时区与经度的选择,应当把它写明。
假说 · 尚未验证
作者的假说(尚未验证): 现行通用的节气日期表(以东八区标准时记日)与按各地视太阳时记日的结果,在中国西部的差异比例与本章模型给出的 8% 到 9% 一致,误差不超过一个百分点。什么结果会证明它错:用高精度历表重算后,西部的日期差异比例与 8% 到 9% 相差一个百分点以上。
小结
开普勒方程用牛顿法三步可解到机器精度。由轨道偏心率产生的一年一周分量(幅度 7.66 分钟)与由黄赤交角产生的一年两周分量(9.86 分钟)叠加成均时差,全年在 −14.2 到 +16.4 分钟之间。节气是时刻,日期是约定:在东八区西部,近十分之一的节气按钟表时间与按当地太阳时间会落在不同的日子;另有约 2% 的节气离午夜太近,低精度历表无法为它们定日。
参考文献
Fan, Y. (2026). Five-Node Cyclic Network Dynamics, Solar-Longitude Ephemeris and Multi-Scale Oscillator Perturbations: Rock-Paper-Scissors-Lizard-Spock Lotka-Volterra Graphs, Equation-of-Time Keplerian Mechanics and Jacobian Sensitivity Bounds [五节点循环网络动力学、黄经历表与多尺度振子扰动] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090168 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).