Charlotte Degot A more accurate way to calculate emissions TED

Transcriber: Leslie Gauthier
Reviewer:

For decades now,

we’ve been saying we should
reduce our emissions,

but they’ve kept increasing.

One of the key reasons
is we don’t measure accurately

the climate impact of our actions.

Imagine trying to save money,

but when you go shopping,

there is no price tag on any item …

or trying to lose weight,

but you cannot measure
the portion sizes and the calories.

You would be bound to fail.

This level of blindness
is close to the one we have

when it comes to our climate impact.

Measuring greenhouse gas
emissions is hard.

It has no color,

it has no smell;

it’s invisible.

We cannot put sensors everywhere,

on every building,

every track,

every field,

every cow –

so most of the time,

we give up and we don’t measure.

And when we do measure,

we are reduced to relying on estimations

and conversion factors.

The consequence is we end up working
with highly incomplete

and inaccurate estimations
of our emissions.

Often we have a margin of error
of 30 to 60 percent.

This means targets
and action plans are set

based on inaccurate data.

If we look at the corporations

that report their progress
on climate to the CDP,

which is a nonprofit organization
that runs a global disclosure system

for environmental impacts,

what we see is striking:

more than two-thirds of the companies

are not accurately measuring
their emissions,

and only seven percent of those companies

are ultimately reducing
their impact in some way.

You cannot reduce
what you cannot measure.

It is key for corporations to be able
to measure across all activities,

all sources that drive carbon up or down.

In a way,

that’s just putting the same rigor
to carbon measurements

that we have for financial accounting.

It took more than 100 years to put modern,
automated financial accounting in place.

We don’t have 100 years
when it comes to climate.

But this is crucial for corporations
to set meaningful targets

and successful action plans.

One of the most powerful tools we have

to help us accelerate on this journey
is artificial intelligence.

Artificial intelligence can process
data automatically

from diverse, unstructured sources

like invoices, consumer behavior data.

It can work by modeling to better estimate
the missing information.

It can simulate and ultimately
optimize emissions.

Let me share an example
of how this could work.

A wine and spirits international company:

billions of sales,

hundreds of brands,

consumers across the globe.

When they want to measure their impact,

they need to measure
across the entire set of their emissions.

This means direct emissions
from facilities,

purchased electricity,

raw materials,

leased assets,

IT emissions

business travel,

transportation,

waste,

product end of life,

etcetera, etcetera.

That’s a huge amount
of information to collect.

And most of it is actually inaccessible
to the company itself

because it comes from outside
its direct scope of activity.

For example,

from suppliers that are not yet able
to calculate their emissions either.

So when the sustainability team
calculates their impact,

they have no choice
but to do rough estimates.

Let’s examine the glass for bottles.

The way they calculate
glass emissions is the following.

They take the total amount
of glass bought last year –

let’s say 1,000 tons.

They multiply it by a conversion factor,

which represents the average kilos
of CO2 equivalent for one ton of glass –

let’s say 950.

950 x 1000 makes 950,000.

Of course this is hugely inaccurate

because it does not take into account

all the numerous factors
that impact actual emissions,

so it’s hard to set targets
and action plans.

This is where the sustainability team
calls data scientists

to come in and process detailed data
about the type of glass,

the color of the glass,

the recycling share,

the supplier country of origin,

the transportation mode,

by brand,

by product.

They can simulate the design
and the supply chain

and integrate in the calculation

the importance of the glass color –

1.5 times more emissions
for a clear bottle

versus a green bottle;

the importance of the country of origin –

twice the amount of emissions
for one country versus another one,

depending on the energy mix;

the importance of the design itself –

for the same total weight,

1.5 times more emissions
for one design versus another one.

Instead of having one big, average number,

you now have a model
which correlates and calculates emissions

at a granular level.

With this type of methodology,

the emissions figure is typically
corrected by 30 to 50 percent.

And more importantly,

the company can now move to action

as they can, one, set meaningful targets,

two, identify very concrete initiatives,

and three,

recalculate emissions over time
and measure their progress.

Let me share another example:

cement.

Cement is a massive CO2 emitter.

If cement were a country,

it would rank as
the third-largest emitter,

right after China and the US,

in front of the European Union and India.

Most of the emissions come from
the process of producing clinker,

the key ingredient in cement.

To produce clinker,

you need to maintain a temperature
of over 1,400 degrees Celsius.

It requires a lot of fuel,

and it’s really just carbon
containing the whole materials.

So the secret sauce is to produce
cleaner and higher quality clinker,

because the higher
the quality of the clinker,

the less of it you will need
to produce cement ultimately,

and therefore the less emissions
you will generate.

But producing high-quality clinker
is a complex science.

It depends on multiple factors
that influence each other.

For example, the process parameters,

like the rotation speed of the machine,

how quickly you fill it,

the type of fuel you use,

the raw materials
and their exact chemical composition.

This is where artificial intelligence
can again have an enormous impact.

On-site operational teams
are trying to manually maintain

the best set of parameters possible.

AI can help by measuring better
through different sources,

like direct measurements,

material and mass balance,

etcetera …

simulate all the potential decisions

and recommend the optimal ones
to the operators.

These techniques implemented
in a cement production process

enable a substantial emissions reduction

in a matter of months.

There is an infinity
of applications possible.

There is no company,

no industry that cannot derive
significant climate impact

from the use of artificial intelligence.

I’m not saying artificial intelligence
alone will save us.

But artificial intelligence,

by helping us measure accurately,

simulate

and optimize,

enables significant emissions reduction

in a quite fast, cheap and easy way.

We cannot miss this opportunity.

Thank you.

抄写员:Leslie Gauthier
审稿人:

几十年来,

我们一直在说我们应该
减少排放,

但它们一直在增加。

关键原因之一
是我们没有准确

衡量我们的行动对气候的影响。

想象一下试图省钱,

但当你去购物时,

任何商品都没有价格标签……

或者试图减肥,

但你无法
测量份量和卡路里。

你注定会失败。

当涉及到我们的气候影响时,这种失明程度接近我们所拥有的程度。

测量温室气体
排放量很困难。

它没有颜色,

没有气味;

它是不可见的。

我们不能将传感器放置在任何地方

、每栋建筑物、

每条赛道、

每块田地、

每头奶牛上——

所以大多数时候,

我们放弃了,我们不进行测量。

当我们进行测量时,

我们只能依赖估计

和转换因子。

结果是我们最终对我们的排放量进行
了高度不完整

和不准确的估计

我们通常有
30% 到 60% 的误差幅度。

这意味着目标
和行动计划是

根据不准确的数据设定的。

如果我们看看

那些
向 CDP 报告气候进展的公司,CDP

是一个
运行全球

环境影响披露系统的非营利组织,

我们看到的是惊人的:

超过三分之二的

公司没有准确衡量
他们的 排放,

而这些公司中只有 7%

最终
以某种方式减少了它们的影响。

你不能减少
你无法衡量的东西。

对于企业而言,
能够衡量所有活动、

所有导致碳上升或下降的来源是关键。

在某种程度上,

这与

我们对财务会计的碳测量一样严格。

实施现代化的自动化财务会计花了 100 多年
的时间。 在气候方面,

我们没有 100 年的时间

但这对于
企业设定有意义的目标

和成功的行动计划至关重要。

我们必须帮助我们加速这一旅程的最强大的工具之一

是人工智能。

人工智能可以
自动处理

来自各种非结构化来源的数据,

如发票、消费者行为数据。

它可以通过建模来更好地
估计缺失的信息。

它可以模拟并最终
优化排放。

让我分享一个
如何工作的例子。

一家葡萄酒和烈酒国际公司:

数十亿的销售额、

数百个品牌、

全球消费者。

当他们想要衡量其影响时,

他们需要衡量
整个排放量。

这意味着
来自设施、

购买的电力、

原材料、

租赁资产、

IT 排放、

商务旅行、

运输、

废物、

产品生命周期结束

等的直接排放。

这是
需要收集的大量信息。

而且其中大部分实际上
是公司本身无法访问的,

因为它来自
其直接活动范围之外。

例如,

来自还
无法计算其排放量的供应商。

因此,当可持续发展团队
计算其影响时,

他们
别无选择,只能进行粗略估计。

让我们检查一下玻璃瓶。

他们计算
玻璃排放量的方式如下。

他们拿走
了去年购买的玻璃总量——

比如说 1,000 吨。

他们将其乘以一个换算系数,该系数

表示一吨玻璃的平均二氧化碳当量公斤数——

比如说

950。950 x 1000 等于 950,000。

当然,这是非常不准确的,

因为它没有考虑到

影响实际排放的所有众多因素,

因此很难设定目标
和行动计划。

这就是可持续发展团队
召集数据

科学家进来并处理
有关玻璃类型、

玻璃颜色

、回收份额

、供应商原产国

、运输方式

、品牌

、产品的详细数据的地方。

他们可以模拟设计
和供应链,

并在计算中整合

玻璃颜色的重要性——

透明瓶子的排放量

是绿色瓶子的 1.5 倍;

原产国的重要性——一个国家

的排放量
是另一个国家的两倍,

具体取决于能源组合;

设计本身的重要性——

对于相同的总重量,

一种设计的排放量是另一种设计的 1.5 倍。 您现在拥有的模型

不再是一个大的平均数字,而是一个在粒度级别

关联和计算排放的模型

使用这种方法

,排放数据通常会
修正 30% 到 50%。

更重要的是

,公司现在可以采取

行动,一是设定有意义的目标,

二是确定非常具体的举措

,三是

随着时间的推移重新计算排放量
并衡量其进展。

让我再举一个例子:

水泥。

水泥是一种巨大的二氧化碳排放物。

如果水泥是一个国家,

它将
成为第三大排放国

,仅次于中国和美国

,排在欧盟和印度之前。

大部分排放来自
生产熟料的过程,熟料

是水泥的关键成分。

要生产熟料,

您需要保持
超过 1,400 摄氏度的温度。

它需要大量燃料,

而且它实际上只是
包含所有材料的碳。

所以秘诀是生产
更清洁、质量更高的熟料,因为熟料

质量越高,

最终生产水泥所需的熟料就越

少,因此产生的排放
也就越少。

但生产优质熟料
是一门复杂的科学。

它取决于
相互影响的多种因素。

例如,工艺参数,

如机器的转速

、填充速度、

使用的燃料类型

、原材料
及其确切的化学成分。

这就是人工智能
可以再次产生巨大影响的地方。

现场运营团队
正试图手动维护

一组可能的最佳参数。

人工智能可以通过不同来源更好地测量来提供帮助

例如直接测量、

材料和质量平衡

等……

模拟所有潜在决策

并向操作员推荐最佳决策。

在水泥生产过程中

实施的这些技术可以

在几个月内大幅减少排放。

有无限
的应用可能。

没有一家公司,

没有一个行业不能从人工智能的使用中获得
显着的气候影响

我并不是说
只有人工智能才能拯救我们。

但人工智能

通过帮助我们准确测量、

模拟

和优化,

能够

以一种相当快速、廉价和简单的方式显着减少排放。

我们不能错过这个机会。

谢谢你。