The problem is a spontaneous shutdown with a random frequency of +- the same time of 21 minutes

[SultanRR]

The problem is a spontaneous shutdown with a random frequency of +- the same time of 21 minutes

esphome:
name: "office_v2"
friendly_name: Wemos Office
esp8266:
board: d1_mini
logger:
level: VERBOSE
logs:
switch: ERROR
api:
ota:
- platform: esphome
wifi:
networks:
- ssid: Pro4
password: Milk33kg
power_save_mode: NONE

ap:
ssid: ProE
password: Milk33kg
binary_sensor:
- platform: gpio
pin: GPIO16
#званок двор пин D0
name: "Alarm_dvor_D0"
filters:
- delayed_on: 50ms # Фильтр дребезга (задержка 50 мс)
- platform: gpio
pin: GPIO12
#геркон двери пин D6
name: "Sens_door_D6"
filters:
- delayed_on: 50ms # Фильтр дребезга (задержка 50 мс)
on_press:
then:
- switch.turn_on: output_alarm
- delay: 0.1s
- switch.turn_off: output_alarm
- platform: gpio
pin:
number: GPIO14 # номер GPIO пина, к которому подключена кнопка
#звонок офис пин D5
mode: INPUT_PULLUP # Включаем внутренний подтягивающий резистор (обычно используется с кнопками, замыкающими на GND)
inverted: true # Инвертируем логику
name: "Alarm_office_D5"
filters:
- delayed_on: 50ms # Фильтр дребезга (задержка 50 мс)
on_press:
then:
- switch.turn_on: output_alarm
- delay: 1s
- switch.turn_off: output_alarm
- platform: gpio
pin:
number: GPIO01
inverted: False
name: "PIR Sensor Office"
device_class: motion
filters:
- delayed_on: 3s
switch:
- platform: gpio
pin: GPIO04
#выход(звонок) пин D2
id: output_alarm
name: "Alarm_D2"
- platform: gpio
pin: GPIO05
#выход(звонок) пин D1
id: output_pinD1
name: "Pin_D1"
- platform: template
name: "Virt_1_Bell"
id: output_1_bell
optimistic: true
- platform: template
name: "Virt_2_Bell"
id: output_2_bell
optimistic: true
- platform: gpio
pin: GPIO2 # D4 (или другой свободный пин)
id: my_led
name: "info"
inverted: True

I don 't know what to do anymore . only when working from a powerbank does it seem to work stably, when working from a usb charger (different), it turns off at random for almost always 21 minutes, it can turn off after 4 hours, or maybe after 26 hours. I tried different wires, the blocks were also different , I put ferrite rings on both the supply wire and the input wires, all the pins on the input and output are tightened to +- or cons in accordance with the operating mode of the nominal value of 2+ kOhms. Moving the access point does not affect , the vemos d1 itself was also changed to a board from another manufacturer (soldered slightly differently ) , the same effect , other vem controllers and others working in this room with the same wifi do not experience such problems with disconnection . At the same time , when hanging for this 21 minutes , if you measure the board , then the 5v before the stabilizer and after 3.3V are stable . The board starts automatically after 21 minutes +- 30 seconds. The logs are empty before this glitch.
Given such a stable shutdown for exactly 21 minutes, I think maybe this is some kind of process that can work like that , or something else .. otherwise, there would have been a quick reconnect after the failure . Maybe I touched (used) a sensitive pin......

[ESP_Sprite]

Hard to say, but I'd try adding a bulk capacitor (e.g. 100uF) over the 5V power supply close to the Wemos module.