关于「相位误差的价钱:报童模型下的订货时点」,《节气、热滞后与季节需求》给出的定量结论是什么?
结论:第六章用报童模型标价:照节气名字订货只拿到最优利润的 0.7196,比完全不考虑季节的平均订货(0.7530)还低,而气候变化造成的相位漂移只损失 0.00147。
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。
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 标准库与结论吻合的打印值: 0.7196, 0.7530
| critical fractile 0.5714, z = 0.1800 | ||||
| assumed lag | error | season profit | share of best | fill rate |
| 0.0 | -32.5 | 41,694 | 0.7196 | 0.7766 |
| 10.0 | -22.5 | 47,923 | 0.8271 | 0.8351 |
| 20.0 | -12.5 | 53,981 | 0.9317 | 0.8880 |
| 27.5 | -5.0 | 57,218 | 0.9876 | 0.9152 |
| 32.5 | +0.0 | 57,938 | 1.0000 | 0.9212 |
| 37.5 | +5.0 | 57,220 | 0.9876 | 0.9150 |
| 44.0 | +11.5 | 54,545 | 0.9414 | 0.8919 |
| 55.0 | +22.5 | 48,131 | 0.8307 | 0.8344 |
| no season at all (flat plan): 43,626, share 0.7530, fill 0.8042 |
查看完整验算程序
# Chapter 6. p61: what a phase error costs a weekly newsvendor
# ILLUSTRATIVE perishable seasonal line (same demand curve as p52:
# 30 + 12.0 E[max(0, T - 18)] units a day, true lag 32.5 days, 26
# weeks from d = -60 after Xiazhi). Weekly demand is normal with mean
# mu and sd 0.25 mu. Price 10, unit cost 6, salvage 3 (made-up), so
# the critical fractile is (10 - 6) / (10 - 3). Each week the buyer
# orders Q = m + z s where m is the forecast under an assumed lag l
# and s = 0.25 m (the newsvendor rule of Arrow, Harris and Marschak
# 1951, see also B-211). Expected sales mu - sigma L(z) with the
# normal loss L(z) = phi(z) - z (1 - Phi(z)); no stock carries over.
import math
YEAR = 365.2422
W = 2 * math.pi / YEAR
PRICE, COST, SALVAGE = 10.0, 6.0, 3.0
CV = 0.25
TRUE_LAG = 32.5
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 inv_cdf(p):
lo, hi = -10.0, 10.0
for _ in range(200):
mid = (lo + hi) / 2
if cdf(mid) < p:
lo = mid
else:
hi = mid
return (lo + hi) / 2
def pos_part(mu, s):
return mu * cdf(mu / s) + s * phi(mu / s)
def weekly(lag, first=-60, weeks=26):
out = []
for w in range(weeks):
days = []
for i in range(7):
d = first + 7 * w + i
m = 14.0 + 14.0 * math.cos(W * (d - lag))
days.append(30 + 12.0 * pos_part(m - 18, 3.0))
out.append(math.fsum(days))
return out
Z = inv_cdf((PRICE - COST) / (PRICE - SALVAGE))
def week_profit(q, mu):
sigma = CV * mu
z = (q - mu) / sigma
sold = mu - sigma * (phi(z) - z * (1 - cdf(z)))
return PRICE * sold + SALVAGE * (q - sold) - COST * q, sold
def season(plan, truth):
prof, sold = [], []
for m, mu in zip(plan, truth):
p, s = week_profit(m * (1 + Z * CV), mu)
prof.append(p)
sold.append(s)
return math.fsum(prof), math.fsum(sold)
if __name__ == "__main__":
truth = weekly(TRUE_LAG)
demand = math.fsum(truth)
best, _ = season(truth, truth)
print("critical fractile {:.4f}, z = {:.4f}".format(
(PRICE - COST) / (PRICE - SALVAGE), Z))
print("assumed lag error season profit share of best fill rate")
for lag in (0.0, 10.0, 20.0, 27.5, 32.5, 37.5, 44.0, 55.0):
p, s = season(weekly(lag), truth)
print("{:11.1f} {:+5.1f} {:13,.0f} {:13.4f} {:9.4f}".format(
lag, lag - TRUE_LAG, p, p / best, s / demand))
flat, s = season([demand / 26] * 26, truth)
print("no season at all (flat plan): {:,.0f}, share {:.4f}, fill {:.4f}"
.format(flat, flat / best, s / demand))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 6 章「相位误差的价钱:报童模型下的订货时点」(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}
}