关于「实证闭环:节气订货日历、门店排期与预注册检验」,《节气、热滞后与季节需求》给出的定量结论是什么?

结论:第八章给出买手订货日历、门店节气排期、蒸馏小工具「节气订货日历」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。

数据口径:书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,订货日期只是演示,不对任何商家承诺销量。 第四章说明「哪天最热」是一个被天气噪声主导的统计量:示意数据中单日最高温的定位误差为 16.81 天,而三年逐日数据的谐波相位误差只有 1.25 天。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库

Solar-Term Order Calendar 2026 (illustrative, lag 32.5 d)
Xiazhi 06-21 16:24Dongzhi 12-22 04:54peak 26-07-24trough 27-01-23
fu: chufu 07-15zhongfu 07-25mofu 08-14out 08-24(40 d)
categoryleadstartsdaysorder bynear termnamedshift
iced drinks2126-06-0310226-05-06Lixia+026-04-03+32.5
fans4526-06-187126-04-27Guyu+726-03-26+32.5
sunscreen3026-04-2112226-03-15Chunfen-526-03-15+0.0
hot pot2126-12-0210226-11-04Lidong-326-10-03+32.5
down coats9026-12-157626-09-09Bailu+226-08-08+32.5
spring jackets6026-02-2211925-12-17Dongzhi-525-11-14+32.5
same calendar with a maritime lag of 44.0 d: iced drinks order by 26-05-17
查看完整验算程序
# Chapter 8. p81: the solar-term order calendar (core of tool.py)
# Functions copied unchanged from distill/tool.py. For 2026: the
# terms (p21 formula, Beijing time), heat peak and cold trough
# (solstice plus a 32.5-day lag, Trenberth 1983), the fu periods
# (p31 geng-day rule) and, per category, the half-peak selling window
# of p53, the latest order date (window start minus a 7-day shelf
# buffer minus the supplier lead time) and the date a buyer reaches
# by reading the season off the term names (lag 0). Thresholds and
# lead times are ILLUSTRATIVE: heat-driven iced drinks (18 C, lead
# 21 d) and fans (24 C, 45 d), sun-driven sunscreen (30 d), cold-
# driven hot pot (10 C, 21 d) and down coats (5 C, 90 d), and spring
# jackets driven by the warming rate (60 d). Temperature: mean 14.0,
# amplitude 14.0, day-to-day sd 3.0 C. "near term" names the closest
# term and the days from it; "shift" is order date minus named date.
# Tool examples in the chapter: --lag 41.3 (the coastal site of p33),
# --mean 16.5 --amp 11.0 --sd 2.5 (a milder, damper city).
import datetime
import math

YEAR = 365.2422
W = 2 * math.pi / YEAR
J2000 = datetime.datetime(2000, 1, 1, 12)
BJ = datetime.timedelta(hours=8)
NAMES = ["Chunfen", "Qingming", "Guyu", "Lixia", "Xiaoman", "Mangzhong",
         "Xiazhi", "Xiaoshu", "Dashu", "Liqiu", "Chushu", "Bailu",
         "Qiufen", "Hanlu", "Shuangjiang", "Lidong", "Xiaoxue", "Daxue",
         "Dongzhi", "Xiaohan", "Dahan", "Lichun", "Yushui", "Jingzhe"]
# name, driver, threshold C, first scan day after Xiazhi, lead days
CATS = [("iced drinks", "heat", 18.0, -120, 21),
        ("fans", "heat", 24.0, -120, 45),
        ("sunscreen", "sun", 0.0, -120, 30),
        ("hot pot", "cold", 10.0, 60, 21),
        ("down coats", "cold", 5.0, 60, 90),
        ("spring jackets", "rise", 0.0, -300, 60)]


def sun_lon(t):
    n = (t - J2000).total_seconds() / 86400
    big_l = (280.460 + 0.9856474 * n) % 360
    g = math.radians((357.528 + 0.9856003 * n) % 360)
    return (big_l + 1.915 * math.sin(g) + 0.020 * math.sin(2 * g)) % 360


def crossing(year, target):
    """Beijing datetime when the Sun's longitude reaches target."""
    def gap(x):
        return (sun_lon(x) - target + 180) % 360 - 180
    t = datetime.datetime(year, 1, 1) - BJ
    day = datetime.timedelta(days=1)
    while not (gap(t) < 0 <= gap(t + day)):
        t += day
    lo, hi = t, t + day
    while (hi - lo).total_seconds() > 1:
        mid = lo + (hi - lo) / 2
        if gap(mid) < 0:
            lo = mid
        else:
            hi = mid
    return hi + BJ


def terms(year):
    out = [(crossing(year, 15 * k), k) for k in range(24)]
    return sorted(out)


def cycle(d):
    return (d.toordinal() + 1721425 + 49) % 60


def geng_from(d, k):
    seen = 0
    while True:
        if cycle(d) % 10 == 6:
            seen += 1
            if seen == k:
                return d
        d += datetime.timedelta(days=1)


def fu_dates(year):
    xz = crossing(year, 90).date()
    lq = crossing(year, 135).date()
    c1, c2, c3 = geng_from(xz, 3), geng_from(xz, 4), geng_from(lq, 1)
    out = c3 + datetime.timedelta(days=10)
    return c1, c2, c3, out, (out - c1).days


def phi(z):
    return math.exp(-0.5 * z * z) / math.sqrt(2 * math.pi)


def cdf(z):
    return 0.5 * (1 + math.erf(z / math.sqrt(2)))


def pos_part(mu, s):
    return mu * cdf(mu / s) + s * phi(mu / s)


def excess(driver, c, d, lag, mean, amp, sd):
    m = mean + amp * math.cos(W * (d - lag))
    if driver == "heat":
        return pos_part(m - c, sd)
    if driver == "cold":
        return pos_part(c - m, sd)
    if driver == "sun":
        return max(0.0, math.cos(W * d))
    return max(0.0, -amp * W * math.sin(W * (d - lag)))


def window(driver, c, first, lag, mean, amp, sd):
    grid = [first + 0.25 * i for i in range(4 * 300)]
    vals = [excess(driver, c, d, lag, mean, amp, sd) for d in grid]
    k = max(range(len(grid)), key=lambda i: vals[i])
    lo = hi = k
    while lo > 0 and vals[lo - 1] >= vals[k] / 2:
        lo -= 1
    while hi < len(grid) - 1 and vals[hi + 1] >= vals[k] / 2:
        hi += 1
    return grid[k], grid[lo], grid[hi]


def nearest(t, table):
    best = min(table, key=lambda x: abs((t - x[0]).total_seconds()))
    days = int(round((t - best[0]).total_seconds() / 86400))
    return "{}{:+d}".format(NAMES[best[1]], days)


def calendar(year=2026, lag=32.5, mean=14.0, amp=14.0, sd=3.0,
             buffer=7, leads=None):
    leads = dict(leads or {})
    xz = crossing(year, 90)
    dz = crossing(year, 270)
    table = terms(year - 1) + terms(year) + terms(year + 1)
    rows = []
    for name, driver, c, first, lead in CATS:
        lead = leads.get(name, lead)
        pk, a, b = window(driver, c, first, lag, mean, amp, sd)
        _, a0, _ = window(driver, c, first, 0.0, mean, amp, sd)
        start = xz + datetime.timedelta(days=a)
        order = start - datetime.timedelta(days=buffer + lead)
        named = xz + datetime.timedelta(days=a0 - buffer - lead)
        rows.append({"category": name, "driver": driver, "lead": lead,
                     "start": start, "end": xz + datetime.timedelta(
                         days=b), "order_by": order, "named": named,
                     "shift": (order - named).total_seconds() / 86400,
                     "term": order})
    return {"year": year, "lag": lag, "xz": xz, "dz": dz,
            "heat_peak": xz + datetime.timedelta(days=lag),
            "cold_trough": dz + datetime.timedelta(days=lag),
            "fu": fu_dates(year), "rows": rows, "table": table,
            "terms": terms(year)}


def lines(cal):
    f = "%y-%m-%d"
    out = ["Solar-Term Order Calendar {} (illustrative, lag {:.1f} d)"
           .format(cal["year"], cal["lag"])]
    out.append("Xiazhi {}  Dongzhi {}  peak {}  trough {}".format(
        cal["xz"].strftime("%m-%d %H:%M"), cal["dz"].strftime("%m-%d %H:%M"),
        cal["heat_peak"].strftime(f), cal["cold_trough"].strftime(f)))
    c1, c2, c3, end, n = cal["fu"]
    out.append("fu: chufu {}  zhongfu {}  mofu {}  out {}  ({} d)".format(
        c1.strftime("%m-%d"), c2.strftime("%m-%d"),
        c3.strftime("%m-%d"), end.strftime("%m-%d"), n))
    out.append("category        lead  starts    days  order by  "
               "near term    named     shift")
    for r in cal["rows"]:
        days = (r["end"] - r["start"]).total_seconds() / 86400
        out.append("{:<15} {:4d}  {}  {:4.0f}  {}  {:<12} {}  {:+5.1f}"
                   .format(r["category"], r["lead"],
                           r["start"].strftime(f), days,
                           r["order_by"].strftime(f),
                           nearest(r["order_by"], cal["table"]),
                           r["named"].strftime(f), r["shift"]))
    return out


if __name__ == "__main__":
    for ln in lines(calendar()):
        print(ln)
    late = calendar(lag=44.0)
    print("same calendar with a maritime lag of 44.0 d: iced drinks order"
          " by {}".format(late["rows"][0]["order_by"].strftime("%y-%m-%d")))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《二十四节气气温滞后相位、三伏一阶热惯性与季节性消费需求预测母典》第 8 章「实证闭环:节气订货日历、门店排期与预注册检验」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-232 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026232,
  author    = {Fan, Yafu},
  title     = {Temperature Phase Lag across the Twenty-Four Solar Terms, First-Order Thermal Inertia behind the Dog Days and Seasonal Demand Forecasting},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论